fix(editor): 移除asset_id min_length限制+order默认None (#1468)

AI Code Review 第三轮修复：
1. EditorClipBatchItem.asset_id 移除 min_length=1，允许空字符串（占位片段场景）
2. EditorClipBatchItem.order 默认值改为 None（非0），仅在显式传入时加入 dict
3. 服务层 order 逻辑：dict 无 order key 时回退到索引 i
4. 测试更新：9 passed

alembic/versions/057_add_title_config_to_generation_tasks.py

@@ -0,0 +1,26 @@
+"""Add title_config to generation_tasks
+
+Revision ID: 057_title_config
+Revises: 056_fix_cover_templates_config
+Create Date: 2026-08-23
+"""
+
+import sqlalchemy as sa
+
+from alembic import op
+
+revision = "057_title_config"
+down_revision = "056_fix_cover_templates_config"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "generation_tasks",
+        sa.Column("title_config", sa.JSON(), nullable=False, server_default="{}"),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("generation_tasks", "title_config")

apps/api/app/api/routes/generation_cover.py

@@ -31,8 +31,6 @@ logger = logging.getLogger(__name__)
 router = APIRouter(tags=["Generation"])
 
 
-
-
 # ── Schemas ──────────────────────────────────────────────────────────────
 
 
@@ -65,6 +63,83 @@ class GenerateCoverResponse(BaseModel):
 # ── Route ────────────────────────────────────────────────────────────────
 
 
+def _persist_cover_frame(
+    frame_url: str,
+    plan_id: str,
+    title_text: str = "",
+    *,
+    title_color: str = "#ffffff",
+    title_position: str = "bottom",
+    title_font_size: int | None = None,
+) -> str:
+    """下载 MediaKit 返回的临时帧图，可选叠加标题后转存到 OSS covers/ 路径。
+
+    Args:
+        frame_url: MediaKit 返回的临时帧图 URL
+        plan_id: 剪辑计划 ID（生成 OSS key）
+        title_text: 非空时用 Pillow 在帧上叠加标题（用于 E2 从源素材抽帧，
+            因为源素材本身没有烧录标题）
+        title_color: 标题字体颜色（#RRGGBB）
+        title_position: 标题位置 top/center/bottom
+        title_font_size: 标题字号，None 时自动计算
+    """
+    import tempfile
+    import uuid
+    from pathlib import Path
+
+    tmp_path: str | None = None
+    try:
+        import httpx
+
+        resp = httpx.get(frame_url, timeout=30, follow_redirects=True)
+        resp.raise_for_status()
+        if not resp.content:
+            return frame_url
+
+        with tempfile.NamedTemporaryFile(suffix=".jpg", delete=False) as tmp:
+            tmp.write(resp.content)
+            tmp_path = tmp.name
+
+        # E2 从源素材抽帧时，源素材无标题，叠加标题文字
+        if title_text and title_text.strip():
+            try:
+                from packages.shared.title_overlay import apply_title_to_image
+
+                applied = apply_title_to_image(
+                    tmp_path,
+                    title_text,
+                    color=title_color,
+                    position=title_position,
+                    font_size=title_font_size,
+                )
+                if applied:
+                    logger.info("[封面生成] E2 帧图已叠加标题: plan_id=%s", plan_id)
+            except Exception:
+                logger.warning(
+                    "[封面生成] E2 标题叠加失败（返回无标题帧）: plan_id=%s",
+                    plan_id,
+                    exc_info=True,
+                )
+
+        from packages.shared.storage import get_shared_storage_service
+
+        storage = get_shared_storage_service()
+        cover_key = f"covers/{plan_id}/cover_{uuid.uuid4().hex[:8]}.jpg"
+        storage.upload_file(
+            file_or_path=tmp_path,
+            storage_key=cover_key,
+            content_type="image/jpeg",
+        )
+        public_url = storage.get_url(cover_key)
+        return public_url or frame_url
+    except Exception:
+        logger.warning("封面帧转存失败，返回原始 URL: plan_id=%s", plan_id, exc_info=True)
+        return frame_url
+    finally:
+        if tmp_path:
+            Path(tmp_path).unlink(missing_ok=True)
+
+
 @router.post("/generate-cover", response_model=GenerateCoverResponse)
 def generate_cover(
     body: GenerateCoverRequest,
@@ -198,44 +273,31 @@ def generate_cover(
                 exc_info=True,
             )
 
-    # 仍然找不到才报 400
-    if not rendered_storage_key:
-        logger.error("[封面生成] ❌ 找不到预览视频: plan_id=%s", plan_id)
-        raise HTTPException(
-            status_code=400,
-            detail="请先生成预览视频，再生成封面",
-        )
-
-    # 回写到 plan.config
-    plan_svc.update_plan_config(plan_id, {"rendered_storage_key": rendered_storage_key})
-
-    # 使用裸 URL（rendered/* 已配置公开读）
+    # 使用裸 URL（rendered/* 已配置公开读）；找不到渲染视频时不立即报错，
+    # 因为步骤 E 可以直接从源素材抽帧（历史数据或 Worker 抽帧失败时的兜底）
     primary_video_url = None
-    try:
-        if rendered_storage_key.startswith("http"):
-            primary_video_url = rendered_storage_key
-        else:
-            from packages.shared.storage import get_shared_storage_service
+    if rendered_storage_key:
+        plan_svc.update_plan_config(plan_id, {"rendered_storage_key": rendered_storage_key})
+        try:
+            if rendered_storage_key.startswith("http"):
+                primary_video_url = rendered_storage_key
+            else:
+                from packages.shared.storage import get_shared_storage_service
 
-            storage_svc = get_shared_storage_service()
-            primary_video_url = storage_svc.get_url(rendered_storage_key)
-        # 防御性规范化：合并路径中的双斜杠（// -> /），但保留协议头的 ://
-        # 历史数据中 project_id 为空时会产生 projects//tasks/ 路径，
-        # MediaKit 的 HTTP 客户端会规范化 URL 导致 404
-        if primary_video_url:
-            import re as _re
+                storage_svc = get_shared_storage_service()
+                primary_video_url = storage_svc.get_url(rendered_storage_key)
+            if primary_video_url:
+                import re as _re
 
-            primary_video_url = _re.sub(r"(?<!:)//", "/", primary_video_url)
-        logger.info(
-            "获取预览视频URL用于封面生成: plan_id=%s url=%s",
-            plan_id,
-            primary_video_url[:80] if primary_video_url else "",
-        )
-    except Exception as e:
-        raise HTTPException(
-            status_code=500,
-            detail=f"获取预览视频URL失败: {e}",
-        ) from e
+                primary_video_url = _re.sub(r"(?<!:)//", "/", primary_video_url)
+            logger.info(
+                "获取预览视频URL用于封面生成: plan_id=%s url=%s",
+                plan_id,
+                primary_video_url[:80] if primary_video_url else "",
+            )
+        except Exception as e:
+            logger.warning("获取预览视频URL失败: plan_id=%s err=%s", plan_id, e)
+            primary_video_url = None
 
     # 统一封面管道：优先从 GenerationTask.cover_url 读取渲染后视频抽帧的封面
     # 多步查找 cover_url，和查找视频 URL 一样的 fallback 逻辑
@@ -310,6 +372,129 @@ def generate_cover(
                     exc_info=True,
                 )
 
+        # 步骤 D：从 plan.config.cover_candidates 读取（Worker 渲染时写入）
+        if not cover_url_from_task:
+            _candidates = (plan.config or {}).get("cover_candidates") or []
+            if isinstance(_candidates, list) and _candidates:
+                _first = _candidates[0]
+                if isinstance(_first, dict):
+                    cover_url_from_task = _first.get("image_url") or _first.get("url") or ""
+                    if cover_url_from_task:
+                        logger.info(
+                            "[封面生成] 统一管道封面(步骤D-cover_candidates): plan_id=%s url=%s",
+                            plan_id,
+                            cover_url_from_task[:80],
+                        )
+
+        # 步骤 E1：如果有已渲染的预览视频 URL 但 cover_url 未持久化（历史数据），
+        # 直接从渲染视频抽帧
+        if not cover_url_from_task and primary_video_url:
+            try:
+                from packages.shared.mediakit_client import get_mediakit_client
+
+                mk_client = get_mediakit_client()
+                if mk_client.is_available:
+                    logger.info(
+                        "[封面生成] 步骤E1-从渲染视频抽帧: plan_id=%s url=%s",
+                        plan_id,
+                        primary_video_url[:80],
+                    )
+                    snapshots = mk_client.extract_frames(
+                        video_url=primary_video_url,
+                        strategy="SpecifiedFrames",
+                        max_frames=1,
+                        poll_interval=2.0,
+                        max_poll_attempts=5,
+                        max_retries=0,
+                    )
+                    if snapshots:
+                        raw = snapshots[0].get("image_url") or snapshots[0].get("url") or ""
+                        if raw:
+                            cover_url_from_task = _persist_cover_frame(raw, plan_id)
+                            logger.info(
+                                "[封面生成] 统一管道封面(步骤E1-rendered-video): plan_id=%s url=%s",
+                                plan_id,
+                                cover_url_from_task[:80],
+                            )
+            except Exception:
+                logger.warning(
+                    "[封面生成] 步骤E1从渲染视频抽帧失败: plan_id=%s",
+                    plan_id,
+                    exc_info=True,
+                )
+
+        # 步骤 E2：当 A/B/C/D/E1 均未命中（如历史预览任务无 cover_url）时，
+        # 直接从用户选择的第一个视频素材中抽取封面帧作为兜底。API 请求内短超时，不阻塞。
+        if not cover_url_from_task and body.asset_ids:
+            from packages.adapters.sqlalchemy_impl.asset_repository import (
+                SQLAlchemyAssetRepository,
+            )
+            from packages.shared.mediakit_client import get_mediakit_client
+            from packages.shared.storage import get_shared_storage_service
+
+            asset_repo = SQLAlchemyAssetRepository(db)
+            storage_svc = get_shared_storage_service()
+            mk_client = get_mediakit_client()
+            # 从 plan.config 读取完整标题样式，E2 从源素材抽帧时叠加（源素材本身无标题）
+            _e2_title_cfg = (plan.config or {}).get("title", {}) or {}
+            if not isinstance(_e2_title_cfg, dict):
+                _e2_title_cfg = {}
+            _e2_title_text = (_e2_title_cfg.get("text", "") or "").strip() if _e2_title_cfg.get("enabled", True) else ""
+            # 读取标题样式：前端可能传 color 或 font_color，都兼容
+            _e2_title_color = _e2_title_cfg.get("color") or _e2_title_cfg.get("font_color") or "#ffffff"
+            _e2_title_position = _e2_title_cfg.get("position", "bottom") or "bottom"
+            _e2_title_font_size = _e2_title_cfg.get("font_size") or _e2_title_cfg.get("size")
+            if mk_client.is_available:
+                for aid in body.asset_ids:
+                    try:
+                        asset = asset_repo.get(aid)
+                        if not asset or asset.file_type != "video":
+                            continue
+                        sk = asset.storage_key or ""
+                        if not sk:
+                            continue
+                        src_url = sk if sk.startswith("http") else storage_svc.get_url(sk)
+                        if not src_url:
+                            continue
+                        logger.info(
+                            "[封面生成] 步骤E-从素材抽帧: plan_id=%s asset_id=%s url=%s",
+                            plan_id,
+                            aid,
+                            src_url[:80],
+                        )
+                        snapshots = mk_client.extract_frames(
+                            video_url=src_url,
+                            strategy="SpecifiedFrames",
+                            max_frames=1,
+                            poll_interval=2.0,
+                            max_poll_attempts=5,
+                            max_retries=0,
+                        )
+                        if snapshots:
+                            raw = snapshots[0].get("image_url") or snapshots[0].get("url") or ""
+                            if raw:
+                                cover_url_from_task = _persist_cover_frame(
+                                    raw,
+                                    plan_id,
+                                    title_text=_e2_title_text,
+                                    title_color=_e2_title_color,
+                                    title_position=_e2_title_position,
+                                    title_font_size=_e2_title_font_size,
+                                )
+                                logger.info(
+                                    "[封面生成] 统一管道封面(步骤E-source-asset): plan_id=%s url=%s",
+                                    plan_id,
+                                    cover_url_from_task[:80],
+                                )
+                                break
+                    except Exception:
+                        logger.warning(
+                            "[封面生成] 步骤E从素材抽帧失败: plan_id=%s asset_id=%s",
+                            plan_id,
+                            aid,
+                            exc_info=True,
+                        )
+
         if cover_url_from_task:
             # 标题已在预览视频渲染时烧录（ASS字幕），封面帧自然包含标题
             cover_data = {
@@ -325,13 +510,13 @@ def generate_cover(
             return GenerateCoverResponse(plan_id=plan_id, cover=cover_data)
 
         logger.warning(
-            "[封面生成] 统一管道未找到 cover_url: plan_id=%s",
+            "[封面生成] 统一管道未找到 cover_url (A/B/C/D均未命中): plan_id=%s",
             plan_id,
         )
         # ai_frame/ai_regenerate 类型必须从渲染管道获取，不再回退到 AI 服务
         raise HTTPException(
             status_code=400,
-            detail="封面尚未生成，请先重新生成预览视频以触发封面自动提取",
+            detail="封面生成失败：未找到可抽帧的视频素材，请确认已上传视频素材后重试",
         )
 
     from packages.shared.ai_service import run_generate_cover

apps/api/app/api/routes/generation_tasks.py

@@ -16,6 +16,7 @@ from app.core.task_enqueue import (
 from app.dependencies import (
     get_asset_library_repository,
     get_asset_repository,
+    get_db_session,
     get_generated_video_repository,
     get_generation_task_repository,
     get_project_repository,
@@ -32,6 +33,7 @@ from app.schemas.generation_task import (
     ListGenerationTasksResponse,
 )
 from fastapi import APIRouter, Depends, HTTPException
+from sqlalchemy.orm import Session
 
 from packages.application import (
     CreateGenerationTaskCommand,
@@ -69,6 +71,7 @@ def _to_generation_task_response(task) -> GenerationTaskResponse:
         output_height=getattr(task, "output_height", 720),
         cover_url=getattr(task, "cover_url", ""),
         custom_title=getattr(task, "custom_title", ""),
+        title_config=getattr(task, "title_config", {}) or {},
         logs=getattr(task, "logs", "[]"),
         status=task.status,
         progress=task.progress,
@@ -142,6 +145,54 @@ def _select_assets_from_library(
     return [a.id for a in ready_video_assets]
 
 
+
+def _writeback_edit_plan_config(
+    plan_id: str,
+    task_id: str,
+    title_config: dict | None,
+    db: Session,
+) -> None:
+    """任务入队成功后，回写 EditPlan.config：generation_task_id + title_config。
+
+    用 merge 方式更新，不整体覆盖 config，避免丢失其他字段。
+    失败只记日志，不影响任务创建。
+    """
+    if not plan_id:
+        return
+    try:
+        from packages.adapters.sqlalchemy_impl.models import EditPlanModel
+
+        plan_model = db.query(EditPlanModel).filter(EditPlanModel.id == plan_id).first()
+        if plan_model is None:
+            logger.warning("[生成任务] 回写plan.config失败: plan不存在 plan_id=%s", plan_id)
+            return
+
+        current_config = plan_model.config if isinstance(plan_model.config, dict) else {}
+        merged = dict(current_config)
+        merged["generation_task_id"] = task_id
+        if title_config:
+            merged["title_config"] = title_config
+        plan_model.config = merged
+        db.commit()
+        logger.info(
+            "[生成任务] 回写plan.config成功: plan_id=%s task_id=%s keys=%s",
+            plan_id,
+            task_id,
+            list(merged.keys()),
+        )
+    except Exception as e:
+        logger.warning(
+            "[生成任务] 回写plan.config异常（不影响任务创建）: plan_id=%s error=%s",
+            plan_id,
+            e,
+            exc_info=True,
+        )
+        try:
+            db.rollback()
+        except Exception:
+            pass
+
+
 def _resolve_project_and_library(
     request: CreateGenerationTaskRequest,
     project_repository: Any,
@@ -187,6 +238,7 @@ def create_generation_task(
     project_repository: Any = Depends(get_project_repository),
     asset_library_repository: Any = Depends(get_asset_library_repository),
     asset_repository: Any = Depends(get_asset_repository),
+    db: Session = Depends(get_db_session),
 ) -> BatchGenerationTaskResponse:
     logger.info(
         "[生成任务] 接收请求: user_id=%s, template_id=%s, asset_count=%d, mode=%s, count=%d",
@@ -304,6 +356,7 @@ def create_generation_task(
                     output_height=request.output_height,
                     cover_url=request.cover_url,
                     custom_title=request.custom_title,
+                    title_config=request.title_config or {},
                 )
             )
             try:
@@ -315,6 +368,15 @@ def create_generation_task(
                     log_task_status=True,
                 ):
                     created_tasks.append(task)
+                    # 只在首个成功任务时回写一次 plan.config，
+                    # 避免批量生成时循环覆盖 generation_task_id
+                    if request.source_edit_plan_id and len(created_tasks) == 1:
+                        _writeback_edit_plan_config(
+                            plan_id=request.source_edit_plan_id,
+                            task_id=task.id,
+                            title_config=request.title_config,
+                            db=db,
+                        )
                 else:
                     failed_tasks.append(task)
             except UserPendingLimitExceeded as _e:

apps/api/app/api/routes/templates_editor/draft.py

@@ -17,6 +17,8 @@ from fastapi import APIRouter, Depends, HTTPException, Query, status
 
 from .dependencies import get_draft_plan_id, get_editor_services
 from .schemas import (
+    EditorClipBatchUpdateRequest,
+    EditorClipBatchUpdateResponse,
     EditorDraftResponse,
     EditorPublishResponse,
     EditorRollbackRequest,
@@ -126,11 +128,7 @@ def list_template_versions(
             clip_count=len(v.clip_configs),
             change_note=v.change_note,
             published_by=v.published_by,
-            created_at=(
-                v.created_at.isoformat()
-                if hasattr(v.created_at, "isoformat")
-                else str(v.created_at)
-            ),
+            created_at=(v.created_at.isoformat() if hasattr(v.created_at, "isoformat") else str(v.created_at)),
         )
         for v in versions
     ]
@@ -162,3 +160,35 @@ def rollback_template(
         new_version=tpl.version,
         clip_count=len(clip_configs),
     )
+
+
+@router.put("/clips", response_model=EditorClipBatchUpdateResponse)
+def batch_update_clips(
+    template_id: str,
+    req: EditorClipBatchUpdateRequest,
+    plan_id: str = Depends(get_draft_plan_id),
+    services: tuple[EditTemplateService, EditPlanService] = Depends(get_editor_services),
+    _: AuthenticatedUser = Depends(get_current_user),
+):
+    """批量替换草稿clips（全量覆盖，用于前端选择素材后同步片段）
+
+    事务保证：清空→创建→标记ready 在同一数据库事务内完成，
+    任何步骤失败时自动回滚，避免数据不一致。
+    """
+    _, plan_svc = services
+    plan_svc.get_plan_or_raise(plan_id)
+
+    clips_data = []
+    for clip_item in req.clips:
+        item = {
+            "asset_id": clip_item.asset_id,
+            "start_time": clip_item.start_time,
+            "duration": clip_item.duration,
+        }
+        if clip_item.order is not None:
+            item["order"] = clip_item.order
+        clips_data.append(item)
+
+    plan_svc.replace_all_clips_transactional(plan_id, clips_data)
+
+    return EditorClipBatchUpdateResponse(plan_id=plan_id, clip_count=len(req.clips))

apps/api/app/api/routes/templates_editor/generation.py

@@ -56,6 +56,7 @@ logger = logging.getLogger(__name__)
 router = APIRouter(tags=["Template Editor"])
 
 
+# DEPRECATED: 前端已改用 /generation/tasks 体系，此路由保留仅供旧版兼容，计划下线
 @router.post("/generate", response_model=EditPlanGenerateResponse)
 def generate_editor_draft(
     template_id: str,
@@ -196,7 +197,7 @@ def generate_editor_draft(
 
         plan_svc.update_plan_config(plan_id, {"generation_task_id": gen_task.id})
         plan_svc.transition_status(plan_id, EditPlanStatus.RENDERING)
-        celery_app.send_task("worker.render_edit_plan", args=[plan_id])
+        celery_app.send_task("worker.generate_video", args=[gen_task.id])
 
         updated_plan = plan_svc.get_plan_or_raise(plan_id)
 
@@ -285,6 +286,7 @@ def _get_task_output_url(task, gen_task_repo, db) -> str:
     return ""
 
 
+# DEPRECATED: 前端已改用 /generation/tasks 体系，此路由保留仅供旧版兼容，计划下线
 @router.get("/generation-status", response_model=EditPlanGenerationStatusResponse)
 def get_editor_generation_status(
     template_id: str,

apps/api/app/api/routes/templates_editor/schemas.py

@@ -46,6 +46,7 @@ class EditPlanGenerationStatusResponse(BaseModel):
 
 class EditPlanGenerateRequest(BaseModel):
     """模板编辑器触发生成请求体"""
+
     title_config: Optional[Dict[str, Any]] = Field(
         default_factory=dict,
         description="标题配置（可选），渲染时烧录到视频中。支持字段: text/font/font_size/font_color/position/bold/stroke/shadow",
@@ -75,12 +76,8 @@ class AIRecommendRequest(BaseModel):
     """AI 推荐片段方案请求体"""
 
     asset_ids: List[str] = Field(default_factory=list, description="素材 ID 列表")
-    editing_mode: str = Field(
-        default="one_take", description="剪辑模式: one_take / pip / voice_over / voice_pip"
-    )
-    target_duration: float = Field(
-        default=30.0, ge=1.0, le=600.0, description="目标时长（秒）"
-    )
+    editing_mode: str = Field(default="one_take", description="剪辑模式: one_take / pip / voice_over / voice_pip")
+    target_duration: float = Field(default=30.0, ge=1.0, le=600.0, description="目标时长（秒）")
 
 
 class AIRecommendClipItem(BaseModel):
@@ -107,8 +104,6 @@ class AIRecommendResponse(BaseModel):
     confidence: float = Field(..., ge=0.0, le=1.0, description="AI 推荐置信度 (0~1)")
 
 
-
-
 # ── BGM ────────────────────────────────────────────────────────────────────
 
 
@@ -224,9 +219,7 @@ class ClipBatchDeleteResponse(BaseModel):
 class ClipsFromAssetsRequest(BaseModel):
     """从素材批量创建片段请求"""
 
-    asset_ids: List[str] = Field(
-        ..., min_length=1, max_length=200, description="素材 ID 列表，按顺序追加到时间线末尾"
-    )
+    asset_ids: List[str] = Field(..., min_length=1, max_length=200, description="素材 ID 列表，按顺序追加到时间线末尾")
     clip_type: str = Field(default="main", description="片段类型，默认 main")
 
 
@@ -501,6 +494,28 @@ class EditorClipUpdateRequest(BaseModel):
     config: Optional[dict[str, Any]] = None
 
 
+class EditorClipBatchItem(BaseModel):
+    """批量更新clips的单个片段"""
+
+    asset_id: str = Field(default="", max_length=100, description="关联素材ID，可为空（占位片段）")
+    start_time: float = Field(default=0.0, ge=0.0)
+    duration: float = Field(default=0.0, ge=0.0)
+    order: Optional[int] = Field(default=None, ge=0, description="排序，None表示按数组顺序")
+
+
+class EditorClipBatchUpdateRequest(BaseModel):
+    """批量替换clips请求（全量覆盖）"""
+
+    clips: List[EditorClipBatchItem] = Field(default_factory=list)
+
+
+class EditorClipBatchUpdateResponse(BaseModel):
+    """批量更新clips响应"""
+
+    plan_id: str
+    clip_count: int
+
+
 class EditorPublishResponse(BaseModel):
     """发布草稿响应"""
 

apps/api/app/schemas/generation_task.py

@@ -33,6 +33,11 @@ class CreateGenerationTaskRequest(BaseModel):
     voice_ids: list[str] = Field(default_factory=list)
     # ── 来源剪辑计划 ──
     source_edit_plan_id: str = ""
+    # ── 标题配置（结构化，优先于 custom_title 纯文本）──
+    title_config: dict | None = Field(
+        default=None,
+        description="标题样式对象，包含 text/font/font_size/font_color/position/bold/stroke/shadow 等。为空时不影响现有行为。",
+    )
     # ── 视频标题 ──
     video_title: str = Field(default="", description="生成视频的标题/名称，为空则使用默认命名")
     # ── 批量生成 ──
@@ -109,6 +114,7 @@ class GenerationTaskResponse(BaseModel):
     output_height: int = 720
     cover_url: str = ""
     custom_title: str = ""
+    title_config: dict = Field(default_factory=dict)
     status: str
     progress: float
     result_count: int

apps/api/app/services/edit_plan_service.py

@@ -371,6 +371,85 @@ class EditPlanService:
         logger.info("删除所有片段: plan_id=%s count=%d", plan_id, count)
         return count
 
+    def replace_all_clips_transactional(
+        self,
+        plan_id: str,
+        clips_data: list[dict],
+    ) -> int:
+        """事务性地替换所有片段：清空→创建→标记ready，单事务保证原子性。
+
+        Args:
+            plan_id: 计划 ID
+            clips_data: 片段数据列表，每项包含 asset_id/start_time/duration/order
+
+        Returns:
+            int: 创建的片段数量
+
+        Raises:
+            Exception: 任何步骤失败时自动回滚
+        """
+        from packages.adapters.sqlalchemy_impl.models import EditPlanClipModel
+
+        db = self._clip_repo.session
+        try:
+            # 1. 清空现有 clips（不 commit）
+            deleted_count = db.query(EditPlanClipModel).filter(EditPlanClipModel.plan_id == plan_id).delete()
+
+            # 2. 批量创建新 clips（不 commit）
+            for i, clip_item in enumerate(clips_data):
+                order = clip_item.get("order") or i
+                clip = EditPlanClip.create(
+                    plan_id=plan_id,
+                    clip_type="main",
+                    order=order,
+                    asset_id=clip_item.get("asset_id", ""),
+                    start_time=clip_item.get("start_time", 0.0),
+                    duration=clip_item.get("duration", 0.0),
+                )
+                model = EditPlanClipModel(
+                    id=clip.id,
+                    plan_id=clip.plan_id,
+                    clip_type=clip.clip_type,
+                    order=clip.order,
+                    asset_id=clip.asset_id,
+                    text_content=clip.text_content,
+                    start_time=clip.start_time,
+                    duration=clip.duration,
+                    transition_effect=clip.transition_effect,
+                    transition_duration=clip.transition_duration,
+                    playback_speed=clip.playback_speed,
+                    status=clip.status.value,
+                    config=clip.config,
+                )
+                db.add(model)
+
+            # 3. 标记有 asset_id 的 clips 为 ready（不 commit）
+            pending_with_asset = (
+                db.query(EditPlanClipModel)
+                .filter(
+                    EditPlanClipModel.plan_id == plan_id,
+                    EditPlanClipModel.status == "pending",
+                    EditPlanClipModel.asset_id != "",
+                )
+                .all()
+            )
+            for m in pending_with_asset:
+                m.status = "ready"
+
+            # 4. 一次性提交
+            db.commit()
+            logger.info(
+                "事务性替换片段: plan_id=%s deleted=%d created=%d",
+                plan_id,
+                deleted_count,
+                len(clips_data),
+            )
+            return len(clips_data)
+        except Exception:
+            db.rollback()
+            logger.exception("事务性替换片段失败: plan_id=%s", plan_id)
+            raise
+
     # ── 片段分割与合并 ──────────────────────────────────────────────────────
 
     def split_clip(self, clip_id: str, split_time: float) -> Dict[str, Any]:

apps/web/e2e/core-generation.spec.ts

@@ -231,10 +231,7 @@ test.describe("Core generation flow", () => {
       (response) => {
         const url = response.url()
         const path = new URL(url).pathname
-        return (
-          response.request().method() === "POST" &&
-          path.endsWith("/generation/tasks")
-        )
+        return response.request().method() === "POST" && path.endsWith("/generation/tasks")
       },
       { timeout: 30_000 },
     )

apps/web/src/api/bgm/bgm.ts

@@ -4,11 +4,20 @@
 import apiClient from "../client"
 import type { BgmPreset, BgmPresetsQuery } from "./types"
 
-/** 获取 BGM 预设列表 */
-export const getBgmPresets = async (params?: BgmPresetsQuery): Promise<BgmPreset[]> => {
+/**
+ * 获取 BGM 预设列表
+ * @param templateId 模板/草稿 ID
+ * @param params 分类/关键词筛选
+ */
+export const getBgmPresets = async (
+  templateId: string,
+  params?: BgmPresetsQuery,
+): Promise<BgmPreset[]> => {
   const searchParams: Record<string, string> = {}
   if (params?.category) searchParams.category = params.category
   if (params?.keyword) searchParams.keyword = params.keyword
-  const res = await apiClient.get("/bgm/presets", { params: searchParams })
+  const res = await apiClient.get(`/templates/${templateId}/editor/bgm/presets`, {
+    params: searchParams,
+  })
   return res.data?.data ?? res.data ?? []
 }

apps/web/src/api/generation/cover.ts

@@ -1,9 +1,22 @@
 import apiClient from "../client"
 
+export interface GenerateCoverTitleConfig {
+  text?: string
+  font?: string
+  font_size?: number
+  font_color?: string
+  position?: string
+  bold?: boolean
+  stroke?: boolean
+  shadow?: boolean
+}
+
 export interface GenerateCoverRequest {
   asset_ids: string[]
   cover_type?: "ai_frame" | "manual" | "upload" | "ai_regenerate"
   frame_time?: number
+  /** 标题样式，用于在封面上叠加标题文字 */
+  title_config?: GenerateCoverTitleConfig
 }
 
 export interface GenerateCoverResponse {

apps/web/src/api/tasks/tasks.ts

@@ -6,6 +6,7 @@ import apiClient from "../client"
 import type {
   CreateGenerationTaskRequest,
   CreateGenerationTaskResponse,
+  GenerationTaskDetail,
   TaskItem,
   TaskListParams,
   TaskListResponse,
@@ -19,6 +20,12 @@ export const createGenerationTask = async (
   return data
 }
 
+/** 获取单个生成任务详情（轮询用） */
+export const getGenerationTask = async (taskId: string): Promise<GenerationTaskDetail> => {
+  const { data } = await apiClient.get<GenerationTaskDetail>(`/generation/tasks/${taskId}`)
+  return data
+}
+
 /** 获取任务列表（支持分页和筛选） */
 export const getTasks = async (params?: TaskListParams): Promise<TaskListResponse> => {
   const { data } = await apiClient.get<TaskListResponse>("/tasks", {

apps/web/src/api/tasks/types.ts

@@ -82,10 +82,12 @@ export interface CreateGenerationTaskRequest {
     stroke?: boolean
     shadow?: boolean
   }
+  /** 关联的草稿 ID（编辑流程数据链路用） */
+  source_edit_plan_id?: string
 }
 
-/** 创建生成任务响应（对齐后端 GenerationTaskResponse） */
-export interface CreateGenerationTaskResponse {
+/** 单个生成任务详情（对齐后端 GenerationTaskResponse） */
+export interface GenerationTaskDetail {
   id: string
   project_id: string
   asset_library_id: string
@@ -95,8 +97,18 @@ export interface CreateGenerationTaskResponse {
   asset_ids: string[]
   title_ids: string[]
   voice_ids: string[]
+  source_edit_plan_id?: string
   status: string
   progress: number
   result_count: number
   error_message: string
+  error_info?: TaskErrorInfo
+  created_at?: string | null
+  updated_at?: string | null
+}
+
+/** 创建生成任务响应（后端返回批量结构 {items, total}） */
+export interface CreateGenerationTaskResponse {
+  items: GenerationTaskDetail[]
+  total: number
 }

apps/web/src/api/template-editor/editPlans.ts

@@ -4,51 +4,29 @@
 import apiClient from "../client"
 import type {
   EditPlan,
-  EditPlanListParams,
-  EditPlanListResponse,
-  CreateEditPlanRequest,
   UpdateEditPlanRequest,
   GenerateResponse,
   GenerationStatusResponse,
   EditPlanGeneration,
   GeneratedVideo,
-  CopyEditPlanRequest,
 } from "./types"
 
-/** 获取模板草稿列表（支持分页和筛选） */
-export async function getEditPlans(params?: EditPlanListParams): Promise<EditPlanListResponse> {
-  const response = await apiClient.get<EditPlanListResponse>("/templates/drafts", {
-    params,
-  })
-  return response.data
-}
-
 /** 获取单个模板草稿 */
 export async function getEditPlan(templateId: string): Promise<EditPlan> {
   const response = await apiClient.get(`/templates/${templateId}/editor`)
   return response.data
 }
 
-/** 创建模板草稿 */
-export async function createEditPlan(data: CreateEditPlanRequest): Promise<EditPlan> {
-  const response = await apiClient.post("/templates/drafts", data)
-  return response.data
-}
-
-/** 更新模板草稿 */
+/** 更新模板草稿（支持传入 AbortSignal 用于自动保存竞态取消） */
 export async function updateEditPlan(
   templateId: string,
   data: UpdateEditPlanRequest,
+  signal?: AbortSignal,
 ): Promise<EditPlan> {
-  const response = await apiClient.put(`/templates/${templateId}/editor`, data)
+  const response = await apiClient.put(`/templates/${templateId}/editor`, data, { signal })
   return response.data
 }
 
-/** 删除模板草稿 */
-export async function deleteEditPlan(templateId: string): Promise<void> {
-  await apiClient.delete(`/templates/${templateId}/editor`)
-}
-
 /** 触发生成 */
 export async function generateEditPlan(templateId: string): Promise<GenerateResponse> {
   const response = await apiClient.post(`/templates/${templateId}/editor/generate`)
@@ -72,20 +50,3 @@ export async function getGenerationTaskResults(taskId: string): Promise<Generate
   const response = await apiClient.get(`/generation/tasks/${taskId}/results`)
   return response.data.items || response.data || []
 }
-
-/** 取消生成任务 */
-export async function cancelGeneration(templateId: string): Promise<void> {
-  await apiClient.post(`/templates/${templateId}/editor/cancel`)
-}
-
-/** 复制模板草稿（含所有片段配置） */
-export async function copyEditPlan(
-  templateId: string,
-  data?: CopyEditPlanRequest,
-): Promise<EditPlan> {
-  const response = await apiClient.post<EditPlan>(
-    `/templates/${templateId}/editor/copy`,
-    data || {},
-  )
-  return response.data
-}

apps/web/src/api/template-editor/index.ts

@@ -15,10 +15,7 @@ export type {
   EditPlanSegment,
   EditPlanConfig,
   EditPlan,
-  CreateEditPlanRequest,
   UpdateEditPlanRequest,
-  EditPlanListParams,
-  EditPlanListResponse,
   GenerateResponse,
   EditPlanGeneration,
   ClipStatusItem,
@@ -37,7 +34,6 @@ export type {
   ClipReorderResponse,
   ClipBatchDeleteResponse,
   ClipsFromAssetsResponse,
-  CopyEditPlanRequest,
   TransitionEffect,
   MediaAsset,
 } from "./types"
@@ -53,17 +49,12 @@ export {
 
 // 模板草稿 CRUD + 生成
 export {
-  getEditPlans,
   getEditPlan,
-  createEditPlan,
   updateEditPlan,
-  deleteEditPlan,
   generateEditPlan,
   getGenerationStatus,
   getEditPlanGenerations,
   getGenerationTaskResults,
-  cancelGeneration,
-  copyEditPlan,
 } from "./editPlans"
 
 // 片段 CRUD + 批量操作

apps/web/src/api/template-editor/types.ts

@@ -118,6 +118,17 @@ export interface EditPlanConfig {
   generate_count?: number
   /** 素材模式 */
   material_mode?: string
+  /** 前端标题设置（Step4 自动保存，与 title_config 字段分离，不影响后端渲染） */
+  title?: {
+    text?: string
+    font?: string
+    font_size?: number
+    color?: string
+    position?: string
+    bold?: boolean
+    stroke?: boolean
+    shadow?: boolean
+  }
   /** 预览视频 URL（封面生成用） */
   rendered_storage_key?: string
   /** 生成任务 ID */

apps/web/src/api/templates/index.ts

@@ -9,8 +9,6 @@ export type {
   TemplateSegment,
   TemplateListParams,
   TemplateListResponse,
-  GenerateFromTemplateRequest,
-  GenerateFromTemplateResponse,
   CopyTemplateResponse,
 } from "./types"
 
@@ -24,5 +22,4 @@ export {
   getTemplate,
   toggleFavoriteTemplate,
   copyTemplate,
-  generateFromTemplate,
 } from "./templates"

apps/web/src/api/templates/templates.ts

@@ -5,8 +5,6 @@
 import apiClient from "../client"
 import type {
   CopyTemplateResponse,
-  GenerateFromTemplateRequest,
-  GenerateFromTemplateResponse,
   TemplateItem,
   TemplateListParams,
   TemplateListResponse,
@@ -45,15 +43,3 @@ export const copyTemplate = async (templateId: string): Promise<CopyTemplateResp
   const response = await apiClient.post<CopyTemplateResponse>(`/templates/${templateId}/copy`)
   return response.data
 }
-
-/** 从模板生成 */
-export const generateFromTemplate = async (
-  templateId: string,
-  data?: GenerateFromTemplateRequest,
-): Promise<GenerateFromTemplateResponse> => {
-  const response = await apiClient.post<GenerateFromTemplateResponse>(
-    `/templates/${templateId}/generate`,
-    data,
-  )
-  return response.data
-}

apps/web/src/pages/editing-planner/components/bgm-selector/index.tsx

@@ -16,9 +16,17 @@ interface BgmSelectorProps {
   onClose: () => void
   config: BgmMixConfig
   onChange: (config: BgmMixConfig) => void
+  /** 模板/草稿 ID，用于请求 BGM 预设 */
+  templateId?: string
 }
 
-const BgmSelector: React.FC<BgmSelectorProps> = ({ open, onClose, config, onChange }) => {
+const BgmSelector: React.FC<BgmSelectorProps> = ({
+  open,
+  onClose,
+  config,
+  onChange,
+  templateId,
+}) => {
   const {
     presets,
     loading,
@@ -30,7 +38,7 @@ const BgmSelector: React.FC<BgmSelectorProps> = ({ open, onClose, config, onChan
     loadPresets,
     handlePreview,
     stopPreview,
-  } = useBgmSelector(open)
+  } = useBgmSelector(open, templateId)
 
   /* ── 选中 BGM ── */
   const handleSelect = useCallback(

apps/web/src/pages/editing-planner/components/bgm-selector/useBgmSelector.ts

@@ -19,7 +19,7 @@ export const CATEGORY_LIST: {
  * BGM 选择器数据与交互 Hook
  * 封装列表加载、搜索、分类筛选、试听播放逻辑
  */
-export function useBgmSelector(open: boolean) {
+export function useBgmSelector(open: boolean, templateId?: string) {
   const [presets, setPresets] = useState<BgmPreset[]>([])
   const [loading, setLoading] = useState(false)
   const [activeCategory, setActiveCategory] = useState<BgmCategory | "all">("all")
@@ -30,19 +30,23 @@ export function useBgmSelector(open: boolean) {
 
   /* ── 加载 BGM 列表 ── */
   const loadPresets = useCallback(async () => {
+    if (!templateId) {
+      setPresets([])
+      return
+    }
     setLoading(true)
     try {
       const params: { category?: string; keyword?: string } = {}
       if (activeCategory !== "all") params.category = activeCategory
       if (keyword.trim()) params.keyword = keyword.trim()
-      const data = await getBgmPresets(params)
+      const data = await getBgmPresets(templateId, params)
       setPresets(data)
     } catch {
       message.error("加载 BGM 列表失败")
     } finally {
       setLoading(false)
     }
-  }, [activeCategory, keyword])
+  }, [activeCategory, keyword, templateId])
 
   useEffect(() => {
     if (open) loadPresets()

apps/web/src/pages/generate/GeneratePage.tsx

@@ -171,6 +171,7 @@ const GeneratePage: React.FC = () => {
     autoSubtitles,
     bgm,
     generateCount,
+    sourceEditPlanId: editPlanId,
   })
 
   /* ================================================================

apps/web/src/pages/generate/components/FrontendPreviewPlayer.tsx

@@ -17,7 +17,7 @@ import {
 import type { AssetItem } from "@/api/assets"
 import type { EditingTemplate } from "@/api/editing-planner"
 import { useSegmentScheduler, type PlaybackSegment } from "../hooks/useSegmentScheduler"
-import { useCanvasPlayer, isWebCodecsSupported } from "../hooks/useCanvasPlayer"
+import { useCanvasPlayer } from "../hooks/useCanvasPlayer"
 
 interface FrontendPreviewPlayerProps {
   assets: AssetItem[]
@@ -82,7 +82,9 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
   titleSettings,
 }) => {
   const segments = useMemo(() => buildPlaybackSegments(assets, template), [assets, template])
-  const useWebCodecs = isWebCodecsSupported()
+  // 默认走原生 video 播放（浏览器硬件解码，独立线程，不阻塞 UI）
+  // WebCodecs 仅在明确需要时启用（保留代码作为兜底）
+  const useWebCodecs = false
 
   // ── 两条路径共用同一个 canvas ref（fallback 路径不使用） ──
   const canvasRef = useRef<HTMLCanvasElement>(null)
@@ -112,12 +114,30 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
     [segments],
   )
 
+  // WebCodecs 解码失败后强制走 video fallback
+  const [forceVideoFallback, setForceVideoFallback] = useState(false)
+
+  const handleCanvasError = useCallback((err: Error) => {
+    console.error("[FrontendPreviewPlayer] Canvas decode Error, switching to video fallback:", err)
+    setForceVideoFallback(true)
+  }, [])
+
   const { state: canvasState, controls: canvasControls } = useCanvasPlayer(
     canvasRef,
-    canvasSegments,
-    useWebCodecs ? canvasTitle : undefined,
+    useWebCodecs && !forceVideoFallback ? canvasSegments : [],
+    useWebCodecs && !forceVideoFallback ? canvasTitle : undefined,
+    handleCanvasError,
+    useWebCodecs && !forceVideoFallback,
   )
 
+  // WebCodecs 报告解码失败时自动切换到 video fallback
+  useEffect(() => {
+    if (canvasState.hasDecodeError && !forceVideoFallback) {
+      console.warn("[FrontendPreviewPlayer] hasDecodeError detected, forcing video fallback")
+      setForceVideoFallback(true)
+    }
+  }, [canvasState.hasDecodeError, forceVideoFallback])
+
   // ── Video 播放器（fallback 路径） ──
   const {
     isPlaying: videoIsPlaying,
@@ -130,12 +150,13 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
     videoRefs,
   } = useSegmentScheduler(segments)
 
-  // 选择哪条路径的状态
-  const isPlaying = useWebCodecs ? canvasState.isPlaying : videoIsPlaying
-  const currentTime = useWebCodecs ? canvasState.currentTime : videoCurrentTime
-  const totalDuration = useWebCodecs ? canvasState.duration : videoTotalDuration
-  const canPlay = useWebCodecs ? canvasState.isReady : videoCanPlay
-  const isBuffering = useWebCodecs ? canvasState.isBuffering : false
+  // 选择哪条路径的状态（WebCodecs 解码失败时强制走 video fallback）
+  const effectiveUseWebCodecs = useWebCodecs && !forceVideoFallback
+  const isPlaying = effectiveUseWebCodecs ? canvasState.isPlaying : videoIsPlaying
+  const currentTime = effectiveUseWebCodecs ? canvasState.currentTime : videoCurrentTime
+  const totalDuration = effectiveUseWebCodecs ? canvasState.duration : videoTotalDuration
+  const canPlay = effectiveUseWebCodecs ? canvasState.isReady : videoCanPlay
+  const isBuffering = effectiveUseWebCodecs ? canvasState.isBuffering : false
 
   // ── 配音音频同步 ──
   const audioRef = useRef<HTMLAudioElement | null>(null)
@@ -172,16 +193,18 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
   }, [isPlaying, currentTime])
 
   // 片段切换时同步音频（仅 fallback 路径需要）
-  const segmentSyncKey = useWebCodecs ? -1 : videoCurrentSegIdx
+  const segmentSyncKey = effectiveUseWebCodecs ? -1 : videoCurrentSegIdx
   useEffect(() => {
     const audio = audioRef.current
     if (!audio || !audio.src || !isPlaying) return
     audio.currentTime = currentTime
-  }, [segmentSyncKey, isPlaying, currentTime])
+    // 注意：不要把 currentTime 放进依赖数组，否则每200ms会重置音频位置导致卡顿
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [segmentSyncKey, isPlaying])
 
   const handleSeekTo = useCallback(
     (time: number) => {
-      if (useWebCodecs) {
+      if (effectiveUseWebCodecs) {
         canvasControls.seek(time)
       } else {
         videoSeekTo(time)
@@ -191,11 +214,11 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
         audio.currentTime = time
       }
     },
-    [useWebCodecs, canvasControls, videoSeekTo],
+    [effectiveUseWebCodecs, canvasControls, videoSeekTo],
   )
 
   const handleTogglePlay = useCallback(() => {
-    if (useWebCodecs) {
+    if (effectiveUseWebCodecs) {
       if (canvasState.isPlaying) {
         canvasControls.pause()
       } else {
@@ -204,7 +227,7 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
     } else {
       videoTogglePlayPause()
     }
-  }, [useWebCodecs, canvasState.isPlaying, canvasControls, videoTogglePlayPause])
+  }, [effectiveUseWebCodecs, canvasState.isPlaying, canvasControls, videoTogglePlayPause])
 
   // ── 进度条拖拽 ──
   const [isDragging, setIsDragging] = useState(false)
@@ -247,32 +270,8 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
 
   const progressPercent = totalDuration > 0 ? (currentTime / totalDuration) * 100 : 0
 
-  // ── Canvas ResizeObserver ──
+  // ── Canvas 容器 ref（保留声明，WebCodecs 兜底路径仍引用） ──
   const canvasContainerRef = useRef<HTMLDivElement>(null)
-  useEffect(() => {
-    if (!useWebCodecs || !canPlay) return
-    const container = canvasContainerRef.current
-    const canvas = canvasRef.current
-    if (!container || !canvas) return
-    // 立即设置一次 canvas 像素分辨率，避免默认 300×150 导致首帧变形
-    const initRect = container.getBoundingClientRect()
-    if (initRect.width > 0 && initRect.height > 0) {
-      const dpr = window.devicePixelRatio || 1
-      canvas.width = initRect.width * dpr
-      canvas.height = initRect.height * dpr
-    }
-    const ro = new ResizeObserver((entries) => {
-      for (const entry of entries) {
-        const { width, height } = entry.contentRect
-        if (width > 0 && height > 0) {
-          canvas.width = width * window.devicePixelRatio
-          canvas.height = height * window.devicePixelRatio
-        }
-      }
-    })
-    ro.observe(container)
-    return () => ro.disconnect()
-  }, [useWebCodecs, canPlay])
 
   // ── 未就绪 ──
   if (!ready || !assets.length) {
@@ -298,6 +297,7 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
 
   // ── 无播放片段 ──
   if (!canPlay) {
+    const showDecodeError = forceVideoFallback && canvasState.hasDecodeError
     return (
       <div
         className="xx-preview-empty"
@@ -316,6 +316,19 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
             <LoadingOutlined style={{ fontSize: 48, color: "#fff", marginBottom: 12 }} spin />
             <p style={{ color: "rgba(255,255,255,0.8)" }}>加载中...</p>
           </>
+        ) : showDecodeError ? (
+          <>
+            <PlayCircleOutlined style={{ fontSize: 48, color: "#ef4444", marginBottom: 12 }} />
+            <p className="xx-preview-empty-title" style={{ color: "rgba(255,255,255,0.9)" }}>
+              视频解码失败
+            </p>
+            <p
+              className="xx-preview-empty-desc"
+              style={{ color: "rgba(255,255,255,0.6)", maxWidth: 300, textAlign: "center" }}
+            >
+              {canvasState.errorMessage || "当前浏览器不支持该视频编码格式，请刷新重试"}
+            </p>
+          </>
         ) : (
           <>
             <PlayCircleOutlined
@@ -332,7 +345,7 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
   return (
     <>
       {/* ── Canvas 渲染层（WebCodecs 路径） ── */}
-      {useWebCodecs && (
+      {effectiveUseWebCodecs && (
         <div
           ref={canvasContainerRef}
           style={{
@@ -353,8 +366,8 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
         </div>
       )}
 
-      {/* ── Video 渲染层（fallback 路径） ── */}
-      {!useWebCodecs &&
+      {/* ── Video 渲染层（默认路径，浏览器原生硬件解码） ── */}
+      {!effectiveUseWebCodecs &&
         segments.map((seg, i) => (
           <video
             key={seg.assetId}
@@ -362,9 +375,7 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
             ref={(el) => {
               videoRefs.current[i] = el
             }}
-            preload={
-              i === videoCurrentSegIdx ? "auto" : i === videoCurrentSegIdx + 1 ? "metadata" : "none"
-            }
+            preload="auto"
             src={seg.videoUrl}
             style={{
               position: "absolute",
@@ -423,7 +434,7 @@ const FrontendPreviewPlayer: React.FC<FrontendPreviewPlayerProps> = ({
           zIndex: 10,
         }}
       >
-        {useWebCodecs ? "Canvas" : `片段 ${videoCurrentSegIdx + 1}/${segments.length}`}
+        {`片段 ${videoCurrentSegIdx + 1}/${segments.length}`}
       </div>
 
       {/* 控制条 */}

apps/web/src/pages/generate/components/GenerateStepContent.tsx

@@ -141,6 +141,7 @@ export const GenerateStepContent: React.FC<GenerateStepContentProps> = (props) =
           onSelectedMaterialsChange={onSelectedMaterialsChange}
           smartSelectedIds={smartSelectedIds}
           onSmartSelectedIdsChange={onSmartSelectedIdsChange}
+          selectedTemplate={selectedTemplate}
         />
       )
     case 3:
@@ -156,6 +157,7 @@ export const GenerateStepContent: React.FC<GenerateStepContentProps> = (props) =
         <Step4TitleSettings
           titleSettings={titleSettings}
           onTitleSettingsChange={onTitleSettingsChange}
+          selectedTemplate={selectedTemplate}
         />
       )
     case 5:
@@ -182,6 +184,7 @@ export const GenerateStepContent: React.FC<GenerateStepContentProps> = (props) =
           duration={duration}
           assetIds={materialMode === "auto" ? smartSelectedIds : selectedMaterials}
           selectedTemplate={selectedTemplate}
+          titleSettings={titleSettings}
         />
       )
     case 7:

apps/web/src/pages/generate/components/PreviewVideoPanel.tsx

@@ -16,6 +16,7 @@ import { LoadingOutlined } from "@ant-design/icons"
 import type { AssetItem } from "@/api/assets"
 import type { EditingTemplate } from "@/api/editing-planner"
 import type { TitleSettings } from "../types"
+import { getFontFamily } from "../constants"
 import FrontendPreviewPlayer from "./FrontendPreviewPlayer"
 
 interface PreviewVideoPanelProps {
@@ -87,7 +88,7 @@ function buildTitleStyle(settings: TitleSettings, containerHeight: number): Reac
       : (Math.min(settings.size, 96) / ASS_VIDEO_HEIGHT) * 400 // fallback
 
   const base: React.CSSProperties = {
-    fontFamily: settings.font || "思源黑体",
+    fontFamily: getFontFamily(settings.font),
     fontSize: `${fontSizePx}px`,
     color: settings.color || "#ffffff",
     fontWeight: settings.bold ? 700 : 400,
@@ -176,7 +177,12 @@ const TitleOverlay: React.FC<{ titleSettings: TitleSettings }> = ({ titleSetting
           position: "absolute",
         }}
       >
-        {displayTitle}
+        {displayTitle.split("/").map((part, i) => (
+          <span key={i}>
+            {i > 0 && <br />}
+            {part}
+          </span>
+        ))}
       </div>
     </div>
   )

apps/web/src/pages/generate/components/Step2MaterialSelect.tsx

@@ -15,6 +15,8 @@ interface Step2MaterialSelectProps {
   onSelectedMaterialsChange: (ids: string[]) => void
   smartSelectedIds: string[]
   onSmartSelectedIdsChange: (ids: string[]) => void
+  /** 当前选中的模板/草稿 ID，用于自动保存 */
+  selectedTemplate?: string
 }
 
 const Step2MaterialSelect: React.FC<Step2MaterialSelectProps> = (props) => {

apps/web/src/pages/generate/components/Step4TitleSettings.tsx

@@ -12,6 +12,8 @@ import AiTitleGenerator from "./title/AiTitleGenerator"
 interface Step4TitleSettingsProps {
   titleSettings: TitleSettings
   onTitleSettingsChange: (settings: TitleSettings) => void
+  /** 当前选中的模板/草稿 ID，用于自动保存 */
+  selectedTemplate?: string
 }
 
 const Step4TitleSettings: React.FC<Step4TitleSettingsProps> = (props) => {

apps/web/src/pages/generate/components/Step5VoiceSelect.tsx

@@ -17,6 +17,16 @@ interface Step5VoiceSelectProps {
 }
 
 /** 格式化时长 mm:ss */
+/** 获取素材实际时长（优先顶层 duration，fallback 到 metadata.duration） */
+const getDuration = (item: AssetItem): number => {
+  return item.duration ?? (item.metadata?.duration as number) ?? 0
+}
+
+/** 获取素材实际文件大小 */
+const getFileSize = (item: AssetItem): number => {
+  return item.file_size ?? (item.metadata?.file_size as number) ?? 0
+}
+
 const formatDuration = (seconds?: number): string => {
   if (!seconds || seconds <= 0) return "00:00"
   const m = Math.floor(seconds / 60)
@@ -89,7 +99,7 @@ const Step5VoiceSelect: React.FC<Step5VoiceSelectProps> = ({
       // 如果启用了时长校验，且配音时长不足
       if (totalVideoDuration > 0) {
         const material = materials.find((m) => m.id === id)
-        if (material && (material.duration || 0) < totalVideoDuration) {
+        if (material && getDuration(material) < totalVideoDuration) {
           setPendingVoiceId(id)
           setDurationWarningOpen(true)
           return
@@ -271,25 +281,24 @@ const Step5VoiceSelect: React.FC<Step5VoiceSelectProps> = ({
                 }}
               >
                 <span style={{ display: "flex", alignItems: "center", gap: 4 }}>
-                  {formatDuration(item.duration)}
-                  {totalVideoDuration > 0 &&
-                    (Number(item.duration) || 0) < Number(totalVideoDuration) && (
-                      <span
-                        style={{
-                          color: "#ff4d4f",
-                          fontSize: 11,
-                          fontWeight: 500,
-                          display: "inline-flex",
-                          alignItems: "center",
-                          gap: 2,
-                        }}
-                      >
-                        <WarningOutlined />
-                        时长不足
-                      </span>
-                    )}
+                  {formatDuration(getDuration(item))}
+                  {totalVideoDuration > 0 && getDuration(item) < Number(totalVideoDuration) && (
+                    <span
+                      style={{
+                        color: "#ff4d4f",
+                        fontSize: 11,
+                        fontWeight: 500,
+                        display: "inline-flex",
+                        alignItems: "center",
+                        gap: 2,
+                      }}
+                    >
+                      <WarningOutlined />
+                      时长不足
+                    </span>
+                  )}
                 </span>
-                <span>{formatFileSize(item.file_size)}</span>
+                <span>{formatFileSize(getFileSize(item))}</span>
               </div>
             </div>
           )
@@ -318,7 +327,9 @@ const Step5VoiceSelect: React.FC<Step5VoiceSelectProps> = ({
           return (
             <p>
               该配音时长（
-              <strong>{pendingMaterial ? formatDuration(pendingMaterial.duration) : "--"}</strong>
+              <strong>
+                {pendingMaterial ? formatDuration(getDuration(pendingMaterial)) : "--"}
+              </strong>
               ）短于视频总时长（
               <strong>{formatDuration(totalVideoDuration)}</strong>
               ），播放时配音可能提前结束，建议选择更长的配音素材。

apps/web/src/pages/generate/components/Step6CoverSettings.tsx

@@ -14,6 +14,8 @@ interface Step6CoverSettingsProps {
   assetIds?: string[]
   /** 当前选中的模板 ID */
   selectedTemplate?: string
+  /** Step4 标题设置，用于预览视频烧录标题 & 封面叠加标题 */
+  titleSettings?: import("../types").TitleSettings
 }
 
 const Step6CoverSettings: React.FC<Step6CoverSettingsProps> = (props) => {
@@ -40,6 +42,7 @@ const Step6CoverSettings: React.FC<Step6CoverSettingsProps> = (props) => {
     duration: props.duration,
     assetIds: props.assetIds,
     selectedTemplate: props.selectedTemplate,
+    titleSettings: props.titleSettings,
   })
 
   const handleAutoGenerate = () => {

apps/web/src/pages/generate/components/title/TitlePresetsGrid.tsx

@@ -2,6 +2,7 @@
  * 标题预设样式网格
  */
 import React from "react"
+import { getFontFamily } from "../../constants"
 
 interface TitlePresetItem {
   key: string
@@ -35,7 +36,7 @@ const TitlePresetsGrid: React.FC<TitlePresetsGridProps> = ({
           >
             <span
               className="xx-title-preset-preview-text"
-              style={{ ...p.previewStyle, ...(fontFamily ? { fontFamily } : {}) }}
+              style={{ ...p.previewStyle, fontFamily: getFontFamily(fontFamily || "思源黑体") }}
             >
               标题
             </span>

apps/web/src/pages/generate/constants.ts

@@ -56,6 +56,21 @@ export const FONT_OPTIONS = [
   "华康俪金黑",
 ]
 
+/* ── 标题字体 CSS font-family 映射（中文显示名 → 浏览器可识别的字体栈） ── */
+export const FONT_FAMILY_MAP: Record<string, string> = {
+  思源黑体: '"Source Han Sans SC", "Noto Sans SC", "PingFang SC", "Microsoft YaHei", sans-serif',
+  思源宋体: '"Source Han Serif SC", "Noto Serif SC", "Songti SC", "SimSun", serif',
+  苹方: '"PingFang SC", -apple-system, "Helvetica Neue", sans-serif',
+  PingFang: '"PingFang SC", -apple-system, "Helvetica Neue", sans-serif',
+  微软雅黑: '"Microsoft YaHei", "PingFang SC", sans-serif',
+  楷体: '"KaiTi", "STKaiti", "DFKai-SB", serif',
+  华康俪金黑: '"华康俪金黑", "DFLiJinHei-W8", "Source Han Sans SC", "Microsoft YaHei", sans-serif',
+}
+
+export function getFontFamily(font: string): string {
+  return FONT_FAMILY_MAP[font] || FONT_FAMILY_MAP["思源黑体"]
+}
+
 /* ── 标题样式预设 ── */
 export const TITLE_PRESETS = [
   {

apps/web/src/pages/generate/generate.css

@@ -2309,6 +2309,8 @@
 .xx-cover-preview-box {
   position: relative;
   aspect-ratio: 9 / 16;
+  max-width: 180px;
+  margin: 0 auto;
   background: var(--bg-tertiary);
   border-radius: var(--radius-md);
   overflow: hidden;

apps/web/src/pages/generate/hooks/generate-video/types.ts

@@ -19,6 +19,8 @@ export interface UseGenerateVideoProps {
   autoSubtitles: boolean
   bgm: boolean
   generateCount: number
+  /** 当前草稿 ID（URL 参数 edit_plan_id，用于后端回写任务关联） */
+  sourceEditPlanId?: string | null
 }
 
 /** 生成阶段 */

apps/web/src/pages/generate/hooks/generate-video/useGenerationPolling.ts

@@ -1,87 +1,146 @@
 import { useRef, useCallback } from "react"
 import { message } from "antd"
-import { getGenerationStatus, getGenerationTaskResults } from "@/api/template-editor"
+import axios from "axios"
+import { getGenerationTask } from "@/api/tasks/tasks"
+import { getGenerationTaskResults } from "@/api/template-editor"
 import { safeExtractError } from "./errorUtils"
 
 interface UseGenerationPollingOptions {
-  templateId: string
   onProgress: (progress: number) => void
   onComplete: (videos: unknown[]) => void
   onFailed: (errorMsg: string) => void
 }
 
+/** 最大连续错误次数（仅对可重试错误），超过后终止轮询 */
+const MAX_RETRYABLE_ERRORS = 10
+/** 获取结果的最大重试次数 */
+const MAX_RESULTS_RETRIES = 3
+
 /**
- * 生成状态轮询 Hook
- * 轮询生成状态，更新进度，处理完成/失败
+ * 生成状态轮询 Hook（v2 — 改用 /generation/tasks/{task_id}）
+ *
+ * 旧版轮询 GET /templates/{id}/editor/generation-status 依赖 plan 维度状态，
+ * 在编辑流程数据链路断裂时拿不到 task_id。新版直接使用 POST /generation/tasks
+ * 返回的 task_id 轮询任务详情，不再依赖 plan。
+ *
+ * 错误处理：
+ * - 4xx（尤其 404）视为不可恢复，立即 onFailed，不再重试
+ * - 5xx / 网络错误重试，最多连续 MAX_RETRYABLE_ERRORS 次
+ * - 任务完成后获取结果失败会重试 MAX_RESULTS_RETRIES 次，仍失败则 onFailed
  */
 export const useGenerationPolling = ({
-  templateId,
   onProgress,
   onComplete,
   onFailed,
 }: UseGenerationPollingOptions) => {
   const progressTimer = useRef<ReturnType<typeof setTimeout>>()
+  const cancelledRef = useRef(false)
 
   const clearTimer = useCallback(() => {
+    cancelledRef.current = true
     if (progressTimer.current) {
       clearTimeout(progressTimer.current)
       progressTimer.current = undefined
     }
   }, [])
 
-  const startPolling = useCallback(() => {
-    const poll = async () => {
+  /** 任务完成后拉取结果列表，带重试 */
+  const fetchResultsWithRetry = useCallback(
+    async (taskId: string, attempt = 0): Promise<unknown[] | null> => {
       try {
-        const data = await getGenerationStatus(templateId)
-
-        if (data.plan_status === "completed") {
-          onProgress(100)
-          // 获取生成的视频结果
-          let videos: unknown[] = []
-          if (data.generation_task_id) {
-            try {
-              videos = await getGenerationTaskResults(data.generation_task_id)
-            } catch (err) {
-              console.error("[获取生成结果失败]", err)
-            }
-          }
-          onComplete(videos)
-          message.success("视频生成完成！")
-          return
+        return await getGenerationTaskResults(taskId)
+      } catch (err) {
+        if (cancelledRef.current) return null
+        console.error(`[获取生成结果失败] 第 ${attempt + 1} 次`, err)
+        if (attempt < MAX_RESULTS_RETRIES - 1) {
+          await new Promise((resolve) => setTimeout(resolve, 1000 * (attempt + 1)))
+          return fetchResultsWithRetry(taskId, attempt + 1)
         }
-        if (data.plan_status === "failed") {
-          const dataAny = data as unknown as Record<string, unknown>
-          const rawMsg =
-            dataAny.error_message ||
-            dataAny.error ||
-            dataAny.message ||
-            (Array.isArray(data.clips)
-              ? (data.clips as { status: string; error_message?: string }[]).find(
-                  (c) => c.status === "failed",
-                )?.error_message
-              : undefined) ||
-            "视频生成失败，请联系管理员或重试"
-          const errorMsg = safeExtractError(rawMsg)
-          console.error("[生成失败] templateId:", templateId, "响应:", data)
-          onFailed(errorMsg)
-          message.error(errorMsg)
-          return
-        }
-
-        const clips = data.clips || []
-        const total = clips.length || 1
-        const done = (clips as { status: string }[]).filter((c) => c.status === "completed").length
-        onProgress(Math.round((done / total) * 100))
-
-        progressTimer.current = setTimeout(poll, 2000)
-      } catch (pollErr) {
-        console.error("[轮询出错] templateId:", templateId, pollErr)
-        progressTimer.current = setTimeout(poll, 3000)
+        return null
       }
-    }
+    },
+    [],
+  )
 
-    progressTimer.current = setTimeout(poll, 2000)
-  }, [templateId, onProgress, onComplete, onFailed])
+  const startPolling = useCallback(
+    (taskId: string) => {
+      cancelledRef.current = false
+      let consecutiveErrors = 0
+
+      const poll = async () => {
+        if (cancelledRef.current) return
+        try {
+          const task = await getGenerationTask(taskId)
+          consecutiveErrors = 0
+
+          if (task.status === "completed") {
+            onProgress(100)
+            const videos = await fetchResultsWithRetry(taskId)
+            if (cancelledRef.current) return
+            if (videos === null) {
+              const errorMsg = "视频已生成，但获取结果列表失败，请稍后在任务列表查看"
+              console.error("[生成结果获取失败] taskId:", taskId)
+              onFailed(errorMsg)
+              message.error(errorMsg)
+              return
+            }
+            onComplete(videos)
+            message.success("视频生成完成！")
+            return
+          }
+
+          if (task.status === "failed" || task.status === "cancelled") {
+            const rawMsg =
+              task.error_info?.error_message ||
+              task.error_message ||
+              (task.status === "cancelled" ? "任务已取消" : "视频生成失败，请联系管理员或重试")
+            const errorMsg = safeExtractError(rawMsg)
+            console.error("[生成失败] taskId:", taskId, "响应:", task)
+            onFailed(errorMsg)
+            message.error(errorMsg)
+            return
+          }
+
+          // pending / waiting / running — 继续轮询
+          const pct = Math.max(0, Math.min(99, Math.round(Number(task.progress) || 0)))
+          onProgress(pct)
+          progressTimer.current = setTimeout(poll, 2000)
+        } catch (pollErr) {
+          if (cancelledRef.current) return
+          console.error("[轮询出错] taskId:", taskId, pollErr)
+
+          // 4xx 不可恢复，立即失败
+          const status = axios.isAxiosError(pollErr) ? pollErr.response?.status : undefined
+          if (status && status >= 400 && status < 500) {
+            const msg =
+              (axios.isAxiosError(pollErr) &&
+                (pollErr.response?.data as { detail?: string; message?: string } | undefined)
+                  ?.detail) ||
+              (axios.isAxiosError(pollErr) &&
+                (pollErr.response?.data as { detail?: string; message?: string } | undefined)
+                  ?.message) ||
+              `查询任务失败 (${status})`
+            const errorMsg = safeExtractError(msg)
+            onFailed(errorMsg)
+            message.error(errorMsg)
+            return
+          }
+
+          consecutiveErrors += 1
+          if (consecutiveErrors >= MAX_RETRYABLE_ERRORS) {
+            const errorMsg = "任务状态查询连续失败，请稍后在任务列表查看结果"
+            onFailed(errorMsg)
+            message.error(errorMsg)
+            return
+          }
+          progressTimer.current = setTimeout(poll, 3000)
+        }
+      }
+
+      progressTimer.current = setTimeout(poll, 1500)
+    },
+    [onProgress, onComplete, onFailed, fetchResultsWithRetry],
+  )
 
   return { startPolling, clearTimer }
 }

apps/web/src/pages/generate/hooks/useCanvasPlayer.ts

@@ -9,8 +9,21 @@ import { createFile } from "mp4box"
 import type { Movie, Sample } from "mp4box"
 
 // ── 常量 ──
-/** 初始化预解码最大帧数（约 2 秒 @30fps），后续帧通过 decodeAroundPosition 按需解码 */
-const MAX_INIT_FRAMES = 60
+
+/**
+ * 规范化 mp4box 提取的 codec 字符串为 WebCodecs 兼容格式
+ * mp4box 返回的 codec 可能包含 mp4box 特有后缀（如 avc1.640028），
+ * WebCodecs 要求标准 ISO BMFF codec string（如 avc1.640028）
+ * 大部分情况下格式一致，但需要做防御性处理
+ */
+function normalizeCodecString(codec: string): string {
+  // mp4box 有时返回带空格的 codec，去掉
+  const trimmed = codec.trim()
+  // HEVC: mp4box 可能返回 hev1.1.6.L93.B0 或 hvc1.1.6.L93.B0
+  // WebCodecs 接受 hev1.x.x 或 hvc1.x.x，两者都可
+  // H.264: mp4box 返回 avc1.640028，WebCodecs 也接受
+  return trimmed
+}
 
 // ── MP4 Box 解析辅助函数 ──
 
@@ -96,7 +109,7 @@ class FrameQueue {
   private frames: FrameEntry[] = []
   private maxSize: number
 
-  constructor(maxSize = 5) {
+  constructor(maxSize = 200) {
     this.maxSize = maxSize
   }
 
@@ -108,24 +121,36 @@ class FrameQueue {
     this.frames.push(entry)
   }
 
-  /** 获取当前时间戳应显示的帧 */
+  /** 获取当前时间戳应显示的帧（二分查找，O(log n)） */
   getCurrentFrame(timestamp: number): VideoFrame | null {
-    let best: FrameEntry | null = null
-    let bestIdx = -1
-    for (let i = 0; i < this.frames.length; i++) {
-      const f = this.frames[i]
-      if (f.pts <= timestamp + 0.01) {
-        best = f
-        bestIdx = i
+    if (this.frames.length === 0) return null
+
+    const target = timestamp + 0.01
+
+    // 找到最后一个 pts <= target 的帧（右边界）
+    let lo = 0,
+      hi = this.frames.length - 1,
+      bestIdx = -1
+    while (lo <= hi) {
+      const mid = (lo + hi) >> 1
+      if (this.frames[mid].pts <= target) {
+        bestIdx = mid
+        lo = mid + 1
+      } else {
+        hi = mid - 1
       }
     }
+
+    if (bestIdx < 0) return null
+
+    // 关闭并移除 bestIdx 之前的所有已播放帧
     for (let i = 0; i < bestIdx; i++) {
       this.frames[i].frame.close()
     }
-    if (bestIdx >= 0) {
-      this.frames = this.frames.slice(bestIdx)
-    }
-    return best?.frame ?? null
+    this.frames.splice(0, bestIdx)
+
+    // 此时 bestIdx 对应帧已在索引 0
+    return this.frames[0]?.frame ?? null
   }
 
   clear() {
@@ -172,6 +197,10 @@ export interface CanvasPlayerState {
   duration: number
   isReady: boolean
   isBuffering: boolean
+  /** WebCodecs 解码失败时为 true，调用方应 fallback 到原生 video 播放 */
+  hasDecodeError: boolean
+  /** 解码错误信息（用于 UI 展示） */
+  errorMessage: string
 }
 
 export interface CanvasPlayerControls {
@@ -209,21 +238,29 @@ export function useCanvasPlayer(
     stroke?: boolean
     shadow?: boolean
   },
+  onError?: (error: Error) => void,
+  enabled: boolean = true,
 ) {
   const [state, setState] = useState<CanvasPlayerState>({
-    hasSupport: isWebCodecsSupported(),
+    hasSupport: enabled && isWebCodecsSupported(),
     isPlaying: false,
     currentTime: 0,
     duration: 0,
     isReady: false,
     isBuffering: false,
+    hasDecodeError: false,
+    errorMessage: "",
   })
 
   // ── 内部引用 ──
   const decoderRef = useRef<VideoDecoder | null>(null)
-  const frameQueueRef = useRef(new FrameQueue(600))
-  /** 已解码的片段索引集合，用于按需解码（先标记防重入，失败时移除允许重试） */
-  const decodedSegmentsRef = useRef(new Set<number>())
+  const frameQueueRef = useRef(new FrameQueue(200))
+  /** 每个片段持久化解码器，避免每次新建导致关键帧错误 */
+  const segmentDecodersRef = useRef(new Map<number, VideoDecoder>())
+  /** 每个片段已送入解码器的 sample 游标（用于续解码） */
+  const segmentSampleCursorRef = useRef(new Map<number, number>())
+  /** 后台补充解码是否正在运行（防重入） */
+  const isFeedingRef = useRef(false)
   /** 解码代数计数器，seek 时递增以作废正在进行的异步解码 */
   const decodeGenerationRef = useRef(0)
   const rafRef = useRef<number>(0)
@@ -234,6 +271,8 @@ export function useCanvasPlayer(
   const videoDimRef = useRef<{ width: number; height: number }>({ width: 0, height: 0 })
   const isDestroyedRef = useRef(false)
   const lastProgressUpdateRef = useRef<number>(0)
+  const onErrorRef = useRef(onError)
+  onErrorRef.current = onError
 
   // 计算总时长
   const totalDuration = segments.reduce((sum, seg) => sum + (seg.endTime - seg.startTime), 0)
@@ -367,7 +406,7 @@ export function useCanvasPlayer(
             globalEndTime: globalStart + (segment.endTime - segment.startTime),
             trackId: videoTrack.id ?? 1,
             timescale: videoTrack.timescale ?? 90000,
-            codec: videoTrack.codec ?? "avc1.42E01E",
+            codec: normalizeCodecString(videoTrack.codec ?? "avc1.42E01E"),
             videoWidth: videoTrack.track_width || 1280,
             videoHeight: videoTrack.track_height || 720,
             description,
@@ -451,87 +490,91 @@ export function useCanvasPlayer(
     [segments, extractCodecDescription],
   )
 
-  // ── 初始化 VideoDecoder 并解码指定片段 ──
-  const decodeSegment = useCallback(
-    async (_buffer: ArrayBuffer, meta: SegmentMeta, maxFrames?: number): Promise<void> => {
-      if (isDestroyedRef.current) return
+  // ── 解码片段的一批帧（使用持久化解码器，支持从断点续解码） ──
+  /**
+   * @param segIdx 片段索引
+   * @param maxFrames 本次最多解码多少帧
+   */
+  const decodeSegmentBatch = useCallback(
+    async (segIdx: number, maxFrames: number = 60): Promise<number> => {
+      if (isDestroyedRef.current) return 0
+      const metas = segmentMetaRef.current
+      const meta = metas[segIdx]
+      if (!meta) return 0
+      const buffer = segmentDataRef.current.get(meta.assetId)
+      if (!buffer) return 0
 
-      let decoderReady = false
+      const gen = decodeGenerationRef.current
+      let decoder = segmentDecodersRef.current.get(segIdx)
+      let cursor = segmentSampleCursorRef.current.get(segIdx) ?? 0
+      const samples = meta.samples
 
-      // 配置解码器（每个片段可能需要不同的 codec/分辨率）
-      const decoder = new VideoDecoder({
-        output: (frame: VideoFrame) => {
-          // 从第一帧获取实际尺寸
-          if (videoDimRef.current.width === 0 || videoDimRef.current.height === 0) {
-            videoDimRef.current = { width: frame.codedWidth, height: frame.codedHeight }
-            console.log(
-              `[useCanvasPlayer] Actual frame size: ${frame.codedWidth}x${frame.codedHeight}`,
-            )
-          }
-          const localTime = frame.timestamp / 1_000_000
-          const globalTime = localTime + meta.globalStartTime
-          frameQueueRef.current.push({
-            frame,
-            pts: globalTime,
-            duration: (frame.duration ?? 0) / 1_000_000,
-          })
-        },
-        error: (e: DOMException) => {
-          console.error("[useCanvasPlayer] Decoder error:", e)
-        },
-      })
-
-      console.log("[useCanvasPlayer] configure:", {
-        codec: meta.codec,
-        description: meta.description,
-        descriptionByteLength: meta.description?.byteLength,
-        videoWidth: meta.videoWidth,
-        videoHeight: meta.videoHeight,
-      })
-
-      try {
-        await decoder.configure({
-          codec: meta.codec,
-          ...(meta.description ? { description: meta.description } : {}),
+      // 如果还没有解码器，新建一个（从关键帧开始，不会报 key frame 错误）
+      if (!decoder || decoder.state === "closed") {
+        decoder = new VideoDecoder({
+          output: (frame: VideoFrame) => {
+            if (videoDimRef.current.width === 0 || videoDimRef.current.height === 0) {
+              videoDimRef.current = { width: frame.codedWidth, height: frame.codedHeight }
+            }
+            const localTime = frame.timestamp / 1_000_000
+            const globalTime = localTime + meta.globalStartTime
+            frameQueueRef.current.push({
+              frame,
+              pts: globalTime,
+              duration: (frame.duration ?? 0) / 1_000_000,
+            })
+          },
+          error: (e: DOMException) => {
+            console.error(`[useCanvasPlayer] Segment ${segIdx} decoder error:`, e)
+            // 重置该片段的解码器和游标，允许重试
+            segmentDecodersRef.current.delete(segIdx)
+            segmentSampleCursorRef.current.set(segIdx, 0)
+          },
         })
-        decoderRef.current = decoder
-        decoderReady = true
-        // 标记缓冲结束，让 UI 开始渲染
+
+        try {
+          await decoder.configure({
+            codec: meta.codec,
+            ...(meta.description ? { description: meta.description } : {}),
+          })
+        } catch (err) {
+          console.error(`[useCanvasPlayer] Segment ${segIdx} configure failed:`, err)
+          const error = err instanceof Error ? err : new Error(String(err))
+          setState((s) => ({
+            ...s,
+            isBuffering: false,
+            hasDecodeError: true,
+            errorMessage: `视频解码失败: ${error.message || "不支持的编解码器"}`,
+          }))
+          onErrorRef.current?.(error)
+          return 0
+        }
+
+        segmentDecodersRef.current.set(segIdx, decoder)
+        cursor = 0
+        // 标记缓冲结束
         setState((s) => ({ ...s, isBuffering: false }))
-      } catch (err) {
-        console.error("[useCanvasPlayer] Decoder configure failed for segment:", err)
-        return
       }
 
-      if (!decoderReady) return
+      if (decoder.state !== "configured") return 0
 
-      // 使用 demuxSegment 中已提取并过滤的 samples（前端切片）
-      const samplesCollected = meta.samples
-      console.log(
-        `[useCanvasPlayer] Segment ${meta.assetId}: ${samplesCollected.length} samples to decode`,
-      )
-      if (samplesCollected.length === 0) {
-        console.warn("[useCanvasPlayer] No samples to decode for segment", meta.assetId)
-        return
-      }
-
-      // 送入解码器
-      let decodedCount = 0
-      let skippedCount = 0
-      let decodeErrors = 0
-      for (const sample of samplesCollected) {
-        if (!sample.data || isDestroyedRef.current) {
-          skippedCount++
+      // 从 cursor 继续喂 sample（流水线批量提交，不 await 单个 decode）
+      let decoded = 0
+      let si = cursor
+      while (si < samples.length && decoded < maxFrames) {
+        if (decodeGenerationRef.current !== gen || isDestroyedRef.current) break
+        if ((decoder.state as string) === "closed") break
+        // 帧队列快满时停止提交（这才是真正的背压）
+        if (frameQueueRef.current.size >= 180) break
+        // 解码器内部队列积压过多时短暂让出线程（阈值64，给硬件足够流水线深度）
+        if (decoder.decodeQueueSize > 64) {
+          await new Promise((r) => setTimeout(r, 5))
           continue
         }
-        if (decoder.state === "closed") break
-        // 初始化阶段限制解码帧数，避免帧缓冲溢出
-        if (maxFrames && decodedCount >= maxFrames) {
-          console.log(
-            `[useCanvasPlayer] Segment ${meta.assetId}: init decode limited to ${maxFrames} frames`,
-          )
-          break
-        }
+
+        const sample = samples[si]
+        si++
+        if (!sample.data) continue
 
         const chunk = new EncodedVideoChunk({
           type: sample.is_sync ? "key" : "delta",
@@ -541,85 +584,73 @@ export function useCanvasPlayer(
         })
 
         try {
-          await decoder.decode(chunk) // 修复：await 捕获异步错误
-          decodedCount++
+          decoder.decode(chunk)
+          decoded++
         } catch (e) {
-          decodeErrors++
-          console.warn(`[useCanvasPlayer] Decode chunk error (${decodeErrors}):`, e)
-          // 连续 3 次解码失败，放弃当前片段
-          if (decodeErrors >= 3) {
-            console.error("[useCanvasPlayer] Too many decode errors, aborting segment")
-            break
-          }
+          console.warn(`[useCanvasPlayer] Segment ${segIdx} decode error:`, e)
+          break
+        }
+      }
+
+      cursor = si
+      segmentSampleCursorRef.current.set(segIdx, cursor)
+      // 等待解码器输出帧（最多500ms）
+      if (decoded > 0 && (decoder.state as string) === "configured") {
+        let waited = 0
+        while (frameQueueRef.current.size < Math.min(decoded, 10) && waited < 500) {
+          await new Promise((r) => setTimeout(r, 20))
+          waited += 20
+          if (isDestroyedRef.current) break
         }
       }
       console.log(
-        `[useCanvasPlayer] Segment ${meta.assetId}: decoded ${decodedCount}, skipped ${skippedCount}, errors ${decodeErrors}, decoder.state=${decoder.state}`,
+        `[useCanvasPlayer] Segment ${segIdx} decoded ${decoded} frames, queue size: ${frameQueueRef.current.size}`,
       )
-
-      // flush 仅在解码器状态正常时执行
-      if (decoder.state === "configured") {
-        try {
-          await decoder.flush()
-          console.log(`[useCanvasPlayer] Segment ${meta.assetId}: flush complete`)
-        } catch (e) {
-          console.warn("[useCanvasPlayer] Decoder flush error:", e)
-        }
-      }
+      return decoded
     },
     [],
   )
 
+  // ── 后台持续补充帧 ──
   /**
-   * 按需解码当前播放位置 ±1 个片段。
-   * 在渲染循环中定期调用，避免一次性解码所有片段导致环形缓冲区溢出丢帧。
-   * 使用"先标记再解码"模式防止并发重复解码，失败时移除标记允许重试。
+   * 根据当前播放时间，确保队列中有足够缓冲
+   * 播放循环每 200ms 调用一次
    */
-  const decodeAroundPosition = useCallback(
+  const feedFrames = useCallback(
     async (currentTime: number) => {
-      const metas = segmentMetaRef.current
-      if (!metas || metas.length === 0) return
+      if (isFeedingRef.current) return
+      isFeedingRef.current = true
+      try {
+        const metas = segmentMetaRef.current
+        if (!metas || metas.length === 0) return
 
-      // 记录当前代数，seek 后代数变化则中止
-      const gen = decodeGenerationRef.current
+        // 队列帧数充足时不解码（目标：保持 >= 80 帧缓冲）
+        if (frameQueueRef.current.size >= 80) return
 
-      let targetIdx = -1
-      let acc = 0
-      for (let i = 0; i < metas.length; i++) {
-        const dur = metas[i].globalEndTime - metas[i].globalStartTime
-        if (currentTime < acc + dur) {
-          targetIdx = i
-          break
+        // 找到当前播放的片段
+        let targetIdx = 0
+        let acc = 0
+        for (let i = 0; i < metas.length; i++) {
+          const dur = metas[i].globalEndTime - metas[i].globalStartTime
+          if (currentTime < acc + dur) {
+            targetIdx = i
+            break
+          }
+          acc += dur
         }
-        acc += dur
-      }
-      if (targetIdx === -1) targetIdx = metas.length - 1
 
-      for (
-        let i = Math.max(0, targetIdx - 1);
-        i <= Math.min(metas.length - 1, targetIdx + 1);
-        i++
-      ) {
-        // seek 已作废当前解码任务
-        if (decodeGenerationRef.current !== gen) return
-        if (decodedSegmentsRef.current.has(i)) continue
-        const meta = metas[i]
-        const buffer = segmentDataRef.current.get(meta.assetId)
-        if (!buffer) continue
-        // 先标记为解码中，防止下一帧渲染时重复发起解码
-        decodedSegmentsRef.current.add(i)
-        try {
-          await decodeSegment(buffer, meta, 300)
-        } catch (e) {
-          // 解码失败则移除标记，允许后续重试
-          decodedSegmentsRef.current.delete(i)
-          console.warn(`[useCanvasPlayer] 按需解码片段 ${i} 失败:`, e)
+        // 依次补充：当前片段 → 下一个片段 → 再下一个
+        for (let offset = 0; offset <= 2; offset++) {
+          const idx = targetIdx + offset
+          if (idx >= metas.length) break
+          if (frameQueueRef.current.size >= 180) break
+          await decodeSegmentBatch(idx, 60)
         }
-        // await 后再次检查代数，seek 期间不更新标记
-        if (decodeGenerationRef.current !== gen) return
+      } finally {
+        isFeedingRef.current = false
       }
     },
-    [decodeSegment],
+    [decodeSegmentBatch],
   )
 
   // ── 标题绘制 ──
@@ -636,7 +667,6 @@ export function useCanvasPlayer(
 
       // 按 "/" 分割为多行（"/" 作为手动换行符）
       const lines = title.text.split(/[/／⁄∕]/)
-      console.log("[drawTitle] 原始标题:", JSON.stringify(title.text), "分割后:", lines)
       const lineHeight = fontSize * 1.3
       const totalHeight = lines.length * lineHeight
 
@@ -748,8 +778,10 @@ export function useCanvasPlayer(
         }
         return s
       })
-      // 按需解码当前 ±1 片段
-      decodeAroundPosition(currentTime)
+      // 后台补充帧：队列不足时自动续解码
+      if (frameQueueRef.current.size < 80) {
+        void feedFrames(currentTime)
+      }
     }
 
     if (currentTime >= totalDuration) {
@@ -758,29 +790,46 @@ export function useCanvasPlayer(
     }
 
     rafRef.current = requestAnimationFrame(renderFrame)
-  }, [canvasRef, totalDuration, titleSettings, drawTitle, computeDrawRect, decodeAroundPosition])
+  }, [canvasRef, totalDuration, titleSettings, drawTitle, computeDrawRect, feedFrames])
 
   // ── 播放控制 ──
   const play = useCallback(async () => {
     if (!state.hasSupport || isDestroyedRef.current) return
 
-    // 重播场景：currentTime 已回到起点但 decodedSegmentsRef 仍有旧标记
-    // 此时 FrameQueue 中旧帧已被淘汰，需清空标记让 decodeAroundPosition 重新解码
-    if (state.currentTime <= 0.1 && decodedSegmentsRef.current.size > 0) {
+    // 重播：必须关闭旧解码器、清空队列、重置游标，从头重新解码
+    if (state.currentTime >= totalDuration - 0.1 || state.currentTime <= 0.1) {
       decodeGenerationRef.current++
-      decodedSegmentsRef.current.clear()
-      // 同步清空帧缓冲，避免旧帧残留导致 getCurrentFrame 返回 null
+      // 关闭所有持久化解码器
+      for (const d of segmentDecodersRef.current.values()) {
+        try {
+          if (d.state !== "closed") d.close()
+        } catch {
+          /* noop */
+        }
+      }
+      segmentDecodersRef.current.clear()
+      segmentSampleCursorRef.current.clear()
       frameQueueRef.current.clear()
+      playStartOffsetRef.current = 0
+      setState((s) => ({ ...s, currentTime: 0 }))
+      // 重新初始化解码
+      const metas = segmentMetaRef.current
+      const initialDecodeCount = Math.min(metas.length, 2)
+      for (let i = 0; i < initialDecodeCount; i++) {
+        await decodeSegmentBatch(i, 60)
+      }
     }
 
     setState((s) => ({ ...s, isPlaying: true }))
     playStartRef.current = performance.now()
-    playStartOffsetRef.current = state.currentTime
+    if (state.currentTime < 0.1) {
+      playStartOffsetRef.current = 0
+    } else {
+      playStartOffsetRef.current = state.currentTime
+    }
     lastProgressUpdateRef.current = 0
     rafRef.current = requestAnimationFrame(renderFrame)
-    // 立即触发一次按需解码，不等渲染循环 200ms 节流
-    decodeAroundPosition(state.currentTime)
-  }, [state.hasSupport, state.currentTime, renderFrame, decodeAroundPosition])
+  }, [state.hasSupport, state.currentTime, totalDuration, renderFrame, decodeSegmentBatch])
 
   const pause = useCallback(() => {
     setState((s) => ({ ...s, isPlaying: false }))
@@ -790,33 +839,61 @@ export function useCanvasPlayer(
   const seek = useCallback(
     async (time: number) => {
       const clampedTime = Math.max(0, Math.min(time, totalDuration))
+      decodeGenerationRef.current++
+      // 关闭所有解码器、清空队列、重置游标
+      for (const d of segmentDecodersRef.current.values()) {
+        try {
+          if (d.state !== "closed") d.close()
+        } catch {
+          /* noop */
+        }
+      }
+      segmentDecodersRef.current.clear()
+      segmentSampleCursorRef.current.clear()
+      frameQueueRef.current.clear()
       setState((s) => ({ ...s, currentTime: clampedTime }))
       playStartOffsetRef.current = clampedTime
       playStartRef.current = performance.now()
-      // seek 时递增解码代数，作废正在进行的异步解码
-      decodeGenerationRef.current++
-      // 清空帧队列（clear 内部会 close 所有帧）+ 清空已解码标记
-      frameQueueRef.current.clear()
-      decodedSegmentsRef.current.clear()
-      await decodeAroundPosition(clampedTime)
+      // 找到 seek 目标片段，从该片段开始解码
+      const metas = segmentMetaRef.current
+      let targetIdx = 0,
+        acc = 0
+      for (let i = 0; i < metas.length; i++) {
+        const dur = metas[i].globalEndTime - metas[i].globalStartTime
+        if (clampedTime < acc + dur) {
+          targetIdx = i
+          break
+        }
+        acc += dur
+      }
+      await decodeSegmentBatch(targetIdx, 60)
+      await decodeSegmentBatch(Math.min(targetIdx + 1, metas.length - 1), 60)
     },
-    [totalDuration, decodeAroundPosition],
+    [totalDuration, decodeSegmentBatch],
   )
 
   const destroy = useCallback(() => {
     isDestroyedRef.current = true
     cancelAnimationFrame(rafRef.current)
 
+    // 关闭所有持久化解码器
+    for (const d of segmentDecodersRef.current.values()) {
+      try {
+        if (d.state !== "closed") d.close()
+      } catch {
+        /* noop */
+      }
+    }
+    segmentDecodersRef.current.clear()
+    segmentSampleCursorRef.current.clear()
+
     if (decoderRef.current && decoderRef.current.state !== "closed") {
       decoderRef.current.close()
     }
 
-    // 递增代数中止进行中的异步解码，清空帧队列（clear 内部 close 所有帧）
-    decodeGenerationRef.current++
     frameQueueRef.current.clear()
     segmentDataRef.current.clear()
     segmentMetaRef.current = []
-    decodedSegmentsRef.current.clear()
   }, [])
 
   // ── 预加载下一个片段的数据 ──
@@ -833,6 +910,7 @@ export function useCanvasPlayer(
 
   // ── 初始化：加载并解码所有片段 ──
   useEffect(() => {
+    if (!enabled) return
     if (!state.hasSupport || segments.length === 0) {
       console.log("[useCanvasPlayer] Skip init:", {
         hasSupport: state.hasSupport,
@@ -842,10 +920,24 @@ export function useCanvasPlayer(
     }
 
     let cancelled = false
-    console.log("[useCanvasPlayer] Init start, segments:", segments.length)
 
     const init = async () => {
-      setState((s) => ({ ...s, isBuffering: true }))
+      // ✅ 关键修复：重置销毁标记，允许新的 init 周期正常工作
+      // destroy() 在 useEffect cleanup 中被调用，将 isDestroyedRef 设为 true
+      // 如果不重置，后续的 loadSegment / decodeSegment 会立即 return
+      isDestroyedRef.current = false
+      // ✅ Strict Mode 修复：init 不再递增 generation
+      // seek() 和 play() 仍保留 generation 递增用于中止异步解码
+      // 重置错误状态，避免上一轮的解码错误影响新的 init 周期
+      setState((s) => ({
+        ...s,
+        isBuffering: true,
+        hasDecodeError: false,
+        errorMessage: "",
+        isReady: false,
+      }))
+
+      console.log("[useCanvasPlayer] Init start v2_DIAG, segments:", segments.length)
 
       // 1. 加载所有片段数据
       for (const seg of segments) {
@@ -887,32 +979,44 @@ export function useCanvasPlayer(
 
       segmentMetaRef.current = metas
 
-      // 3. 按需解码：初始只解码前 3 个片段，后续通过 decodeAroundPosition 动态加载
-      //    避免一次性全量解码导致 frameQueue 环形缓冲区旧帧被丢弃引发黑屏
-      decodedSegmentsRef.current.clear()
-      const initGen = decodeGenerationRef.current
-      const initialDecodeCount = Math.min(metas.length, 3)
-      for (let i = 0; i < initialDecodeCount; i++) {
-        if (cancelled) break
-        // seek 或 destroy 已作废当前初始化
-        if (decodeGenerationRef.current !== initGen) break
-        const meta = metas[i]
-        const buffer = segmentDataRef.current.get(meta.assetId)
-        if (!buffer) continue
-        // 先标记为解码中，防止重复解码
-        decodedSegmentsRef.current.add(i)
+      // 3. 初始化解码：关闭旧解码器，前 2 个片段各解 60 帧
+      //    后续由 feedFrames 后台补充
+      for (const d of segmentDecodersRef.current.values()) {
         try {
-          await decodeSegment(buffer, meta, MAX_INIT_FRAMES)
-        } catch (e) {
-          // 解码失败则移除标记，允许后续重试
-          decodedSegmentsRef.current.delete(i)
-          console.warn(`[useCanvasPlayer] 初始化解码片段 ${i} 失败:`, e)
+          if (d.state !== "closed") d.close()
+        } catch {
+          /* noop */
         }
       }
+      segmentDecodersRef.current.clear()
+      segmentSampleCursorRef.current.clear()
+      frameQueueRef.current.clear()
+
+      const initGen = decodeGenerationRef.current
+      const initialDecodeCount = Math.min(metas.length, 2)
+      console.log(
+        `[useCanvasPlayer] Starting init decode: ${initialDecodeCount} segments, metas: ${metas.length}`,
+      )
+      for (let i = 0; i < initialDecodeCount; i++) {
+        if (cancelled) break
+        if (decodeGenerationRef.current !== initGen) break
+        console.log(`[DIAG_v2] Init decode segment ${i}...`)
+        try {
+          await decodeSegmentBatch(i, 60)
+          console.log(`[DIAG_v2] Init decode segment ${i} done`)
+        } catch (e) {
+          console.warn(`[useCanvasPlayer] 初始化解码片段 ${i} 失败:`, e)
+        }
+        if (cancelled) break
+      }
+
+      console.log(`[useCanvasPlayer] Init decode finished, cancelled:`, cancelled)
 
       if (!cancelled) {
-        console.log("[useCanvasPlayer] Init complete, isReady = true")
+        console.log("[useCanvasPlayer] Init complete, isReady = true, duration:", totalDuration)
         setState((s) => ({ ...s, duration: totalDuration, isReady: true, isBuffering: false }))
+      } else {
+        console.warn("[useCanvasPlayer] Init was cancelled before completion")
       }
     }
 

apps/web/src/pages/generate/hooks/useDraftAutoSave.ts

@@ -0,0 +1,124 @@
+/**
+ * 草稿自动保存工具 Hook
+ *
+ * 背景：后端 PUT /templates/{id}/editor 的 config 是「整体替换」语义，
+ * 直接发送 { config: { asset_ids } } 会把 title 等其他字段覆盖掉。
+ * 本 Hook 统一执行「GET 当前 config → 浅合并新字段 → PUT 回去」，
+ * 并用串行队列 + AbortController 保证：
+ *   - 同一时刻只有一个保存请求在飞
+ *   - 快速连续变化时只提交最后一次
+ *   - 组件卸载时取消未完成请求
+ *   - 保存失败时保留补丁，自动重试（指数退避，最多 5 次）
+ *
+ * 保存失败只 console.warn，不弹窗、不阻塞。
+ */
+import { useCallback, useEffect, useRef } from "react"
+import { getEditPlan, updateEditPlan } from "@/api/template-editor"
+
+type ConfigPatch = Record<string, unknown>
+
+/** 最大自动重试次数 */
+const MAX_RETRIES = 5
+/** 初始重试延迟（ms），每次翻倍 */
+const BASE_RETRY_DELAY = 1000
+
+export function useDraftAutoSave(templateId?: string) {
+  const timerRef = useRef<ReturnType<typeof setTimeout>>()
+  const abortRef = useRef<AbortController | null>(null)
+  // 待合并的补丁队列（解决「保存进行中又来了新变化」）
+  const pendingPatchRef = useRef<ConfigPatch | null>(null)
+  const savingRef = useRef(false)
+  const templateIdRef = useRef(templateId)
+  templateIdRef.current = templateId
+
+  const flush = useCallback(async (retryCount = 0) => {
+    const tid = templateIdRef.current
+    if (!tid) return
+    // 已有保存在飞：把新补丁暂存，等当前请求结束后再合并一次
+    if (savingRef.current) return
+
+    // 快照当前补丁，但先不清空 —— 成功后才清除，失败时保留以便重试
+    const patchToSave = pendingPatchRef.current
+    if (!patchToSave) {
+      savingRef.current = false
+      return
+    }
+    savingRef.current = true
+
+    const controller = new AbortController()
+    abortRef.current = controller
+    try {
+      // 1. 读当前 config（拿最新，避免覆盖别人/别的步骤写入的字段）
+      const current = await getEditPlan(tid)
+      if (controller.signal.aborted) return
+      const merged = { ...(current.config || {}), ...patchToSave }
+      // 2. 写回完整合并后的 config
+      await updateEditPlan(tid, { config: merged }, controller.signal)
+      // 3. 保存成功才清除已保存的补丁
+      //    （保存期间可能有新补丁进来，只清除我们已经保存的部分）
+      pendingPatchRef.current = null
+    } catch (err) {
+      const name = (err as { name?: string })?.name
+      if (name === "CanceledError" || name === "AbortError") {
+        // 组件卸载或新请求取消，不重试
+        return
+      }
+      console.warn("[useDraftAutoSave] 自动保存草稿失败:", err)
+
+      // 保存失败：把本次尝试保存的补丁合并回 pendingPatchRef
+      // （保存期间可能有新补丁，新补丁优先）
+      pendingPatchRef.current = {
+        ...patchToSave,
+        ...(pendingPatchRef.current || {}),
+      }
+
+      // 指数退避重试
+      if (retryCount < MAX_RETRIES && !controller.signal.aborted) {
+        const delay = BASE_RETRY_DELAY * Math.pow(2, retryCount)
+        timerRef.current = setTimeout(() => {
+          void flush(retryCount + 1)
+        }, delay)
+      }
+      // 超过最大重试次数后，补丁仍保留在 pendingPatchRef 中，
+      // 下次 scheduleSave 触发时会一起带上
+    } finally {
+      savingRef.current = false
+      // 保存期间又积累了新变化（且不是在重试路径中），再触发一次
+      if (pendingPatchRef.current && !controller.signal.aborted && retryCount === 0) {
+        timerRef.current = setTimeout(() => {
+          void flush()
+        }, 50)
+      }
+    }
+  }, [])
+
+  /**
+   * 调度一次自动保存（防抖）
+   * @param patch 要合并进 config 的局部字段
+   * @param delay 防抖毫秒数
+   */
+  const scheduleSave = useCallback(
+    (patch: ConfigPatch, delay = 500) => {
+      const tid = templateIdRef.current
+      if (!tid) return
+      if (timerRef.current) clearTimeout(timerRef.current)
+      // 累计补丁（同一周期内多次变化合并成一次写入）
+      pendingPatchRef.current = { ...(pendingPatchRef.current || {}), ...patch }
+      timerRef.current = setTimeout(() => {
+        void flush()
+      }, delay)
+    },
+    [flush],
+  )
+
+  useEffect(() => {
+    return () => {
+      if (timerRef.current) clearTimeout(timerRef.current)
+      if (abortRef.current) abortRef.current.abort()
+    }
+  }, [])
+
+  return { scheduleSave }
+}
+
+export default useDraftAutoSave

apps/web/src/pages/generate/hooks/useGenerateVideo.ts

@@ -34,7 +34,6 @@ export function useGenerateVideo(props: UseGenerateVideoProps) {
   }, [])
 
   const { startPolling, clearTimer } = useGenerationPolling({
-    templateId: selectedTemplate,
     onProgress: handleProgress,
     onComplete: handleComplete,
     onFailed: handleFailed,
@@ -90,16 +89,20 @@ export function useGenerateVideo(props: UseGenerateVideoProps) {
       const assetIds =
         props.materialMode === "auto" ? props.smartSelectedIds : props.selectedMaterials
 
+      // 封面 URL：优先 AI 生成缩略图，兜底用户上传
+      const coverUrl = props.coverSettings?.thumbnail_url || props.coverSettings?.upload_url || ""
+
       // 直接创建正式生成任务
-      await createGenerationTask({
+      const taskResp = await createGenerationTask({
         template_id: selectedTemplate,
         asset_ids: assetIds,
         output_width: outputWidth,
         output_height: outputHeight,
-        cover_url: props.coverSettings?.upload_url || "",
+        cover_url: coverUrl,
         custom_title: props.titleSettings?.title || "",
         duration: props.duration || undefined,
         video_ratio: props.videoRatio,
+        ...(props.sourceEditPlanId ? { source_edit_plan_id: props.sourceEditPlanId } : {}),
         ...(props.titleSettings?.title
           ? {
               title_config: {
@@ -116,7 +119,12 @@ export function useGenerateVideo(props: UseGenerateVideoProps) {
           : {}),
       })
 
-      startPolling()
+      // 从创建响应直接拿 task_id，改用新接口轮询
+      const taskId = taskResp.items?.[0]?.id
+      if (!taskId) {
+        throw new Error("创建任务成功但未返回任务 ID，请稍后在任务列表查看")
+      }
+      startPolling(taskId)
     } catch (err: unknown) {
       console.error("[handleGenerate] 生成失败:", err)
       setGenerating(false)

apps/web/src/pages/generate/hooks/useSegmentScheduler.ts

@@ -1,55 +1,37 @@
 /**
- * 素材片段调度器 Hook（多 video 元素方案 v2）
- * 每个片段对应一个独立 <video> 元素，全部预加载，通过 display 切换实现无缝播放
- * 替代单 video + 切 src 方案，消除片段切换延迟
+ * 素材片段调度器 Hook（多 video 元素方案 v3）
+ *
+ * v3 修复：
+ * - 所有动态状态存入 ref，tick 为稳定函数，彻底消除 RAF 闭包陷阱
+ * - 片段切换时先启动下一个 video 再切可见性，消除冻屏间隔
+ * - 进度更新 200ms 节流
  */
+
 import { useState, useRef, useCallback, useEffect, useMemo } from "react"
 
-/** 单个播放片段 */
 export interface PlaybackSegment {
-  /** 素材 ID */
   assetId: string
-  /** 素材视频 URL */
   videoUrl: string
-  /** 片段在素材中的入点（秒） */
   startTime: number
-  /** 片段在素材中的出点（秒） */
   endTime: number
-  /** 片段在时间线中的顺序 */
   order: number
 }
 
-/** 调度器返回 */
 export interface SegmentSchedulerState {
-  /** 是否正在播放 */
   isPlaying: boolean
-  /** 当前播放的全局时间（秒） */
   currentTime: number
-  /** 总时长（秒） */
   totalDuration: number
-  /** 当前片段索引 */
   currentSegmentIndex: number
-  /** 当前片段的本地播放时间 */
   segmentLocalTime: number
-  /** 是否已播完 */
   isEnded: boolean
-  /** 是否可以播放（至少有 1 个片段） */
   canPlay: boolean
-  /** 播放 */
   play: () => void
-  /** 暂停 */
   pause: () => void
-  /** 切换播放/暂停 */
   togglePlayPause: () => void
-  /** 跳转到全局时间 */
   seekTo: (time: number) => void
-  /** 每个片段对应的 video 元素 ref 数组 */
   videoRefs: React.MutableRefObject<(HTMLVideoElement | null)[]>
 }
 
-/**
- * 根据全局时间定位对应的片段和本地时间
- */
 function findSegmentAtTime(
   segments: PlaybackSegment[],
   globalTime: number,
@@ -66,9 +48,6 @@ function findSegmentAtTime(
   return { index: segments.length - 1, localTime: segments[segments.length - 1].endTime }
 }
 
-/**
- * 计算每个片段的全局起始时间
- */
 function buildTimeline(segments: PlaybackSegment[]): number[] {
   const starts: number[] = []
   let acc = 0
@@ -79,224 +58,242 @@ function buildTimeline(segments: PlaybackSegment[]): number[] {
   return starts
 }
 
-/**
- * useSegmentScheduler — 多 video 元素版素材片段调度器
- *
- * 核心改变：
- * - 每个片段对应一个独立 <video> 元素（由组件渲染，ref 传入）
- * - 所有 video 在挂载时即设置 src + preload="auto"，浏览器自动预加载
- * - 切换片段仅改 currentSegmentIndex + display，无需重新 load
- * - 实现无缝切换，无加载延迟
- */
 export function useSegmentScheduler(segments: PlaybackSegment[]): SegmentSchedulerState {
-  /** 每个片段对应的 video 元素 ref（由组件 JSX 渲染并绑定） */
   const videoRefs = useRef<(HTMLVideoElement | null)[]>([])
   const [isPlaying, setIsPlaying] = useState(false)
   const [currentTime, setCurrentTime] = useState(0)
   const [currentSegmentIndex, setCurrentSegmentIndex] = useState(0)
   const [isEnded, setIsEnded] = useState(false)
-  const rafRef = useRef<number>(0)
+  const rafRef = useRef(0)
   const isSeekingRef = useRef(false)
+  const lastTimeUpdateRef = useRef(0)
 
-  // 计算时间线
-  const timelineStarts = useMemo(() => buildTimeline(segments), [segments])
-  const totalDuration = useMemo(
+  // 所有动态值存入 ref，tick 始终读取最新值，不依赖闭包
+  const segIdxRef = useRef(0)
+  const segmentsRef = useRef(segments)
+  const timelineStartsData = useMemo(() => buildTimeline(segments), [segments])
+  const totalDurationData = useMemo(
     () => segments.reduce((sum, seg) => sum + (seg.endTime - seg.startTime), 0),
     [segments],
   )
+  const timelineStartsRef = useRef(timelineStartsData)
+  const totalDurationRef = useRef(totalDurationData)
+  const isPlayingRef = useRef(false)
+
+  segmentsRef.current = segments
+  timelineStartsRef.current = timelineStartsData
+  totalDurationRef.current = totalDurationData
 
   const canPlay = segments.length > 0
 
-  // 当前片段信息
-  const currentSegment = segments[currentSegmentIndex] || null
-  const segmentLocalTime = currentSegment
-    ? currentTime - (timelineStarts[currentSegmentIndex] || 0) + currentSegment.startTime
-    : 0
+  useEffect(() => {
+    segIdxRef.current = currentSegmentIndex
+  }, [currentSegmentIndex])
+
+  useEffect(() => {
+    isPlayingRef.current = isPlaying
+  }, [isPlaying])
+
+  const waitForReady = useCallback((video: HTMLVideoElement, timeout = 3000): Promise<void> => {
+    if (video.readyState >= 3) return Promise.resolve()
+    return new Promise((resolve) => {
+      const onCanPlay = () => {
+        video.removeEventListener("canplay", onCanPlay)
+        clearTimeout(timer)
+        resolve()
+      }
+      const timer = setTimeout(() => {
+        video.removeEventListener("canplay", onCanPlay)
+        resolve()
+      }, timeout)
+      video.addEventListener("canplay", onCanPlay)
+    })
+  }, [])
 
-  /**
-   * 切换到指定片段
-   * 不改变 src（video 已在 JSX 中设置），仅 seek + 等待可播
-   */
   const switchToSegment = useCallback(
-    (index: number, seekToLocalTime?: number): Promise<void> => {
-      return new Promise((resolve) => {
-        // 暂停当前视频
-        const prevVideo = videoRefs.current[currentSegmentIndex]
-        if (prevVideo) prevVideo.pause()
+    async (index: number, seekToLocalTime?: number) => {
+      const segs = segmentsRef.current
+      const video = videoRefs.current[index]
+      if (!video || index >= segs.length) return
 
-        const video = videoRefs.current[index]
-        if (!video || index >= segments.length) {
-          resolve()
-          return
-        }
+      const seg = segs[index]
+      const localTime = seekToLocalTime ?? seg.startTime
+      const oldIdx = segIdxRef.current
+      const oldVideo = videoRefs.current[oldIdx]
 
-        const seg = segments[index]
-        const localTime = seekToLocalTime ?? seg.startTime
+      if (oldVideo && oldVideo !== video) oldVideo.pause()
 
-        // 设置播放位置
+      if (!video.src && seg.videoUrl) {
+        video.src = seg.videoUrl
+        video.load()
+      }
+
+      if (Math.abs(video.currentTime - localTime) > 0.05) {
         video.currentTime = localTime
+      }
 
-        // 如果已有足够帧数据，直接 resolve
-        if (video.readyState >= 2) {
-          setCurrentSegmentIndex(index)
-          resolve()
-          return
-        }
+      segIdxRef.current = index
+      setCurrentSegmentIndex(index)
 
-        // 等待 canplay 事件
-        const onCanPlay = () => {
-          video.removeEventListener("canplay", onCanPlay)
-          clearTimeout(timeoutId)
-          setCurrentSegmentIndex(index)
-          resolve()
-        }
-
-        // 10 秒超时保护
-        const timeoutId = setTimeout(() => {
-          video.removeEventListener("canplay", onCanPlay)
-          console.warn(
-            `[useSegmentScheduler] 片段 ${index} 预加载超时 (10s), readyState=${video.readyState}`,
-          )
-          setCurrentSegmentIndex(index)
-          resolve()
-        }, 10000)
-
-        video.addEventListener("canplay", onCanPlay)
-      })
+      await waitForReady(video)
     },
-    [segments, currentSegmentIndex],
+    [waitForReady],
   )
 
-  /** 播放循环 — 检测片段边界并切换 */
+  // 稳定的 tick 函数，空依赖，所有值从 ref 读取
   const tick = useCallback(() => {
-    const video = videoRefs.current[currentSegmentIndex]
+    const segs = segmentsRef.current
+    const idx = segIdxRef.current
+    const video = videoRefs.current[idx]
+
     if (!video || isSeekingRef.current) {
       rafRef.current = requestAnimationFrame(tick)
       return
     }
 
-    const seg = segments[currentSegmentIndex]
+    const seg = segs[idx]
     if (!seg) return
 
-    // 检查是否到达出点（容差 0.15s）
-    if (video.currentTime >= seg.endTime - 0.15) {
-      video.pause()
-      const nextIndex = currentSegmentIndex + 1
-      if (nextIndex < segments.length) {
-        switchToSegment(nextIndex).then(() => {
-          setIsPlaying(true)
-          rafRef.current = requestAnimationFrame(tick)
-          const nextVideo = videoRefs.current[nextIndex]
-          if (nextVideo) {
-            const canPlay = () => {
-              nextVideo
-                .play()
-                .catch((e) =>
-                  console.warn("[useSegmentScheduler] auto-play next segment failed:", e),
-                )
-            }
-            if (nextVideo.readyState >= 3) {
-              canPlay()
-            } else {
-              const timeout = setTimeout(canPlay, 300)
-              nextVideo.addEventListener(
-                "canplay",
-                () => {
-                  clearTimeout(timeout)
-                  canPlay()
-                },
-                { once: true },
-              )
-            }
+    // 预加载下一个片段
+    const nextIndex = idx + 1
+    if (nextIndex < segs.length) {
+      const nextVideo = videoRefs.current[nextIndex]
+      if (nextVideo) {
+        const timeToEnd = seg.endTime - video.currentTime
+        if (timeToEnd <= 2 && nextVideo.readyState < 3) {
+          const nextSeg = segs[nextIndex]
+          if (Math.abs(nextVideo.currentTime - nextSeg.startTime) > 0.5) {
+            nextVideo.currentTime = nextSeg.startTime
           }
-        })
+        }
+      }
+    }
+
+    // 检测片段边界
+    if (video.currentTime >= seg.endTime - 0.1) {
+      if (nextIndex < segs.length) {
+        const nextVideo = videoRefs.current[nextIndex]
+        const nextSeg = segs[nextIndex]
         const accumulatedTime =
-          (timelineStarts[currentSegmentIndex] || 0) + (seg.endTime - seg.startTime)
+          (timelineStartsRef.current[idx] || 0) + (seg.endTime - seg.startTime)
+
+        if (nextVideo) {
+          if (Math.abs(nextVideo.currentTime - nextSeg.startTime) > 0.1) {
+            nextVideo.currentTime = nextSeg.startTime
+          }
+          // 先启动下一个视频（muted，可安全同时播放）
+          nextVideo
+            .play()
+            .catch((e) => console.warn("[useSegmentScheduler] next segment play failed:", e))
+        }
+
+        // 立即切换可见性
+        segIdxRef.current = nextIndex
+        setCurrentSegmentIndex(nextIndex)
         setCurrentTime(accumulatedTime)
+        lastTimeUpdateRef.current = 0
+        setIsPlaying(true)
+
+        // 下一帧暂停旧视频（让新视频先渲染，避免冻屏）
+        const oldVideo = video
+        requestAnimationFrame(() => {
+          oldVideo.pause()
+        })
+
+        rafRef.current = requestAnimationFrame(tick)
+        return
       } else {
+        video.pause()
         setIsPlaying(false)
         setIsEnded(true)
-        setCurrentTime(totalDuration)
+        setCurrentTime(totalDurationRef.current)
         return
       }
-    } else {
-      const globalTime =
-        (timelineStarts[currentSegmentIndex] || 0) + (video.currentTime - seg.startTime)
-      setCurrentTime(Math.max(0, Math.min(globalTime, totalDuration)))
+    }
+
+    const globalTime = (timelineStartsRef.current[idx] || 0) + (video.currentTime - seg.startTime)
+    const now = performance.now()
+    if (now - lastTimeUpdateRef.current >= 200) {
+      lastTimeUpdateRef.current = now
+      setCurrentTime(Math.max(0, Math.min(globalTime, totalDurationRef.current)))
     }
 
     rafRef.current = requestAnimationFrame(tick)
-  }, [segments, currentSegmentIndex, timelineStarts, totalDuration, switchToSegment])
+  }, [])
 
-  /** 播放 */
   const play = useCallback(async () => {
     if (!canPlay) return
-
     setIsEnded(false)
-
-    // 确保第一段可播放
-    const firstVideo = videoRefs.current[0]
-    if (firstVideo && currentSegmentIndex === 0 && firstVideo.readyState < 2) {
-      await switchToSegment(0)
-    }
-
-    const video = videoRefs.current[currentSegmentIndex]
+    const idx = segIdxRef.current
+    const video = videoRefs.current[idx]
     if (!video) return
 
-    try {
-      const playPromise = video.play()
-      if (playPromise !== undefined) {
-        await playPromise
+    if (idx === 0 && video.readyState < 2) {
+      if (!video.src && segmentsRef.current[0]?.videoUrl) {
+        video.src = segmentsRef.current[0].videoUrl
+        video.load()
       }
+      await waitForReady(video)
+    }
+
+    try {
+      await video.play()
       setIsPlaying(true)
+      cancelAnimationFrame(rafRef.current)
       rafRef.current = requestAnimationFrame(tick)
     } catch (err) {
       console.warn("[useSegmentScheduler] 播放失败:", err)
     }
-  }, [canPlay, switchToSegment, tick, currentSegmentIndex])
+  }, [canPlay, waitForReady, tick])
 
-  /** 暂停 */
   const pause = useCallback(() => {
-    const video = videoRefs.current[currentSegmentIndex]
+    const video = videoRefs.current[segIdxRef.current]
     if (video) video.pause()
     setIsPlaying(false)
     cancelAnimationFrame(rafRef.current)
-  }, [currentSegmentIndex])
+  }, [])
 
-  /** 切换播放/暂停 */
   const togglePlayPause = useCallback(() => {
-    if (isPlaying) {
+    if (isPlayingRef.current) {
       pause()
     } else {
       if (isEnded) {
-        // 播放结束后再次播放，从头开始
         setIsEnded(false)
-        switchToSegment(0, segments[0]?.startTime).then(() => {
-          const video = videoRefs.current[0]
-          if (video) {
-            video.play().catch((e) => console.warn("[useSegmentScheduler] restart play failed:", e))
-            setIsPlaying(true)
-            setCurrentTime(0)
-            rafRef.current = requestAnimationFrame(tick)
-          }
-        })
+        lastTimeUpdateRef.current = 0
+        const firstVideo = videoRefs.current[0]
+        if (firstVideo) {
+          videoRefs.current.forEach((v, i) => {
+            if (v && i !== 0) v.pause()
+          })
+          firstVideo.currentTime = segmentsRef.current[0]?.startTime || 0
+          segIdxRef.current = 0
+          setCurrentSegmentIndex(0)
+          setCurrentTime(0)
+          firstVideo
+            .play()
+            .then(() => {
+              setIsPlaying(true)
+              cancelAnimationFrame(rafRef.current)
+              rafRef.current = requestAnimationFrame(tick)
+            })
+            .catch((e) => console.warn("[useSegmentScheduler] restart failed:", e))
+        }
       } else {
         play()
       }
     }
-  }, [isPlaying, isEnded, pause, play, switchToSegment, segments, tick])
+  }, [isEnded, pause, play, tick])
 
-  /** 跳转到指定全局时间 */
   const seekTo = useCallback(
     async (time: number) => {
       if (!canPlay) return
-
-      const clampedTime = Math.max(0, Math.min(time, totalDuration))
-      const { index, localTime } = findSegmentAtTime(segments, clampedTime)
+      const clampedTime = Math.max(0, Math.min(time, totalDurationRef.current))
+      const { index, localTime } = findSegmentAtTime(segmentsRef.current, clampedTime)
 
       isSeekingRef.current = true
+      cancelAnimationFrame(rafRef.current)
 
-      if (index !== currentSegmentIndex) {
+      if (index !== segIdxRef.current) {
         await switchToSegment(index, localTime)
       } else {
         const video = videoRefs.current[index]
@@ -305,48 +302,54 @@ export function useSegmentScheduler(segments: PlaybackSegment[]): SegmentSchedul
 
       setCurrentTime(clampedTime)
       setIsEnded(false)
+      lastTimeUpdateRef.current = 0
+
+      if (isPlayingRef.current) {
+        const video = videoRefs.current[index]
+        if (video) {
+          video.play().catch(() => {})
+        }
+        rafRef.current = requestAnimationFrame(tick)
+      }
 
       setTimeout(() => {
         isSeekingRef.current = false
       }, 200)
     },
-    [canPlay, totalDuration, segments, currentSegmentIndex, switchToSegment],
+    [canPlay, switchToSegment, tick],
   )
 
-  // 确保 videoRefs 数组长度与 segments 一致 + 强制预加载
   useEffect(() => {
     videoRefs.current = videoRefs.current.slice(0, segments.length)
     while (videoRefs.current.length < segments.length) {
       videoRefs.current.push(null)
     }
-    // 强制预加载：所有 video 元素挂载后，调用 load() 确保浏览器真正开始加载数据
-    videoRefs.current.forEach((video) => {
-      if (video) {
-        video.load()
-      }
-    })
   }, [segments])
 
-  // 组件卸载时清理
   useEffect(() => {
     return () => {
       cancelAnimationFrame(rafRef.current)
     }
   }, [])
 
-  // 片段列表变化时重置
   useEffect(() => {
     cancelAnimationFrame(rafRef.current)
+    segIdxRef.current = 0
     setIsPlaying(false)
     setCurrentTime(0)
     setCurrentSegmentIndex(0)
     setIsEnded(false)
   }, [segments])
 
+  const currentSegment = segments[currentSegmentIndex] || null
+  const segmentLocalTime = currentSegment
+    ? currentTime - (timelineStartsRef.current[currentSegmentIndex] || 0) + currentSegment.startTime
+    : 0
+
   return {
     isPlaying,
     currentTime,
-    totalDuration,
+    totalDuration: totalDurationData,
     currentSegmentIndex,
     segmentLocalTime,
     isEnded,

apps/web/src/pages/generate/hooks/useStep2Materials.ts

@@ -6,6 +6,7 @@ import { useCallback, useEffect, useRef } from "react"
 import { formatDuration } from "../utils/formatDuration"
 import { useMaterialLibrary } from "./step2-materials/useMaterialLibrary"
 import { useSmartMatch } from "./step2-materials/useSmartMatch"
+import { useDraftAutoSave } from "./useDraftAutoSave"
 
 interface UseStep2MaterialsProps {
   materialMode: "manual" | "auto"
@@ -14,6 +15,8 @@ interface UseStep2MaterialsProps {
   onSelectedMaterialsChange: (ids: string[]) => void
   smartSelectedIds: string[]
   onSmartSelectedIdsChange: (ids: string[]) => void
+  /** 当前选中的模板/草稿 ID，用于自动保存 */
+  selectedTemplate?: string
 }
 
 export function useStep2Materials({
@@ -23,6 +26,7 @@ export function useStep2Materials({
   onSelectedMaterialsChange,
   smartSelectedIds,
   onSmartSelectedIdsChange,
+  selectedTemplate,
 }: UseStep2MaterialsProps) {
   const { libraries, selectedLibraryId, setSelectedLibraryId, materials, materialsLoading } =
     useMaterialLibrary()
@@ -53,6 +57,14 @@ export function useStep2Materials({
     handleSmartMatch()
   }, [selectedLibraryId, materialMode, materialsLoading, materials.items, handleSmartMatch])
 
+  /* ── Step2 选择素材后自动保存草稿（防抖 500ms，失败静默） ── */
+  const { scheduleSave } = useDraftAutoSave(selectedTemplate)
+  useEffect(() => {
+    if (!selectedTemplate) return
+    const ids = materialMode === "auto" ? smartSelectedIds : selectedMaterials
+    scheduleSave({ asset_ids: ids }, 500)
+  }, [selectedTemplate, materialMode, selectedMaterials, smartSelectedIds, scheduleSave])
+
   /* ── 手动选择素材 ── */
   const handleToggleMaterial = useCallback(
     (materialId: string) => {

apps/web/src/pages/generate/hooks/useStep4Title/index.ts

@@ -4,17 +4,24 @@ import { getTitles } from "@/api/titles"
 import type { TitleSettings } from "../../types"
 import { useAiTitleGenerator } from "./useAiTitleGenerator"
 import { useTitleStyleUpdaters } from "./useTitleStyleUpdaters"
+import { useDraftAutoSave } from "../useDraftAutoSave"
 
 interface UseStep4TitleProps {
   titleSettings: TitleSettings
   onTitleSettingsChange: (settings: TitleSettings) => void
+  /** 当前选中的模板/草稿 ID，用于自动保存 */
+  selectedTemplate?: string
 }
 
 /**
  * Step 4 标题设置 Hook
  * 封装 AI 标题生成、标题样式设置等逻辑
  */
-export function useStep4Title({ titleSettings, onTitleSettingsChange }: UseStep4TitleProps) {
+export function useStep4Title({
+  titleSettings,
+  onTitleSettingsChange,
+  selectedTemplate,
+}: UseStep4TitleProps) {
   // 标题库数据
   const { data: userTitles = [] } = useQuery({
     queryKey: ["titles"],
@@ -42,6 +49,38 @@ export function useStep4Title({ titleSettings, onTitleSettingsChange }: UseStep4
   const prevAiAutoSelect = useRef(titleSettings.aiAutoSelect)
   const isFirstMount = useRef(true)
 
+  /* ── Step4 标题内容/样式变化后自动保存草稿（防抖 800ms，失败静默） ── */
+  const { scheduleSave: scheduleTitleSave } = useDraftAutoSave(selectedTemplate)
+  useEffect(() => {
+    if (!selectedTemplate) return
+    scheduleTitleSave(
+      {
+        title: {
+          text: titleSettings.title,
+          font: titleSettings.font,
+          font_size: titleSettings.size,
+          color: titleSettings.color,
+          position: titleSettings.position,
+          bold: titleSettings.bold,
+          stroke: titleSettings.stroke,
+          shadow: titleSettings.shadow,
+        },
+      },
+      800,
+    )
+  }, [
+    selectedTemplate,
+    titleSettings.title,
+    titleSettings.font,
+    titleSettings.size,
+    titleSettings.color,
+    titleSettings.position,
+    titleSettings.bold,
+    titleSettings.stroke,
+    titleSettings.shadow,
+    scheduleTitleSave,
+  ])
+
   // 当 AI 自动选择开关打开时，自动生成/选择一个标题填入
   // 首次挂载时如果开关已经是 true 且无标题，也需要触发
   useEffect(() => {

apps/web/src/pages/generate/hooks/useStep6Cover.ts

@@ -7,6 +7,8 @@ import { message } from "antd"
 import type { CoverConfig, CoverTemplate } from "../types/cover"
 import { generateCover } from "@/api/generation"
 import { createPreview, getPreviewStatus } from "@/api/generation/preview"
+import { updateEditPlan } from "@/api/template-editor"
+import type { TitleSettings } from "../types"
 import {
   fetchCoverTemplates,
   createCoverTemplate,
@@ -22,6 +24,8 @@ interface UseStep6CoverProps {
   assetIds?: string[]
   /** 当前选中的模板 ID */
   selectedTemplate?: string
+  /** Step4 标题设置，用于预览视频烧录标题 & 封面叠加标题 */
+  titleSettings?: TitleSettings
 }
 
 export function useStep6Cover({
@@ -30,6 +34,7 @@ export function useStep6Cover({
   duration,
   assetIds = [],
   selectedTemplate = "",
+  titleSettings,
 }: UseStep6CoverProps) {
   const [generating, setGenerating] = useState(false)
 
@@ -92,6 +97,20 @@ export function useStep6Cover({
       const response = await generateCover(selectedTemplate, {
         asset_ids: assetIds,
         cover_type: "ai_frame",
+        ...(titleSettings?.title
+          ? {
+              title_config: {
+                text: titleSettings.title,
+                font: titleSettings.font,
+                font_size: titleSettings.size,
+                font_color: titleSettings.color,
+                position: titleSettings.position,
+                bold: titleSettings.bold,
+                stroke: titleSettings.stroke,
+                shadow: titleSettings.shadow,
+              },
+            }
+          : {}),
       })
       clearTimeout(timeoutId)
       const thumbnailUrl = response.cover?.image_url || ""
@@ -136,6 +155,20 @@ export function useStep6Cover({
             template_id: selectedTemplate,
             asset_ids: assetIds,
             duration: duration || 30,
+            ...(titleSettings?.title
+              ? {
+                  title_config: {
+                    text: titleSettings.title,
+                    font: titleSettings.font,
+                    font_size: titleSettings.size,
+                    font_color: titleSettings.color,
+                    position: titleSettings.position,
+                    bold: titleSettings.bold,
+                    stroke: titleSettings.stroke,
+                    shadow: titleSettings.shadow,
+                  },
+                }
+              : {}),
           })
           // 轮询等待预览渲染完成：递归 setTimeout 避免请求重叠 + 120s 超时兜底
           await new Promise<void>((resolve, reject) => {
@@ -154,6 +187,21 @@ export function useStep6Cover({
               try {
                 const status = await getPreviewStatus(previewResp.task_id)
                 if (status.status === "completed") {
+                  // 保存预览视频地址到 plan.config.rendered_storage_key，
+                  // 供封面 API 的 E1 兜底路径定位渲染后的视频（含标题烧录）。
+                  // video_url 可能是完整 http(s) URL 或 OSS storage_key，两种格式后端都能处理。
+                  if (status.video_url) {
+                    try {
+                      await updateEditPlan(selectedTemplate, {
+                        config: { rendered_storage_key: status.video_url },
+                      })
+                    } catch (saveErr) {
+                      console.warn(
+                        "[Step6] 保存 rendered_storage_key 失败（不阻塞封面重试）:",
+                        saveErr,
+                      )
+                    }
+                  }
                   done(() => resolve())
                 } else if (status.status === "failed") {
                   done(() => reject(new Error(status.error_message || "预览渲染失败")))
@@ -171,6 +219,20 @@ export function useStep6Cover({
           const retryResp = await generateCover(selectedTemplate, {
             asset_ids: assetIds,
             cover_type: "ai_frame",
+            ...(titleSettings?.title
+              ? {
+                  title_config: {
+                    text: titleSettings.title,
+                    font: titleSettings.font,
+                    font_size: titleSettings.size,
+                    font_color: titleSettings.color,
+                    position: titleSettings.position,
+                    bold: titleSettings.bold,
+                    stroke: titleSettings.stroke,
+                    shadow: titleSettings.shadow,
+                  },
+                }
+              : {}),
           })
           const retryUrl = retryResp.cover?.image_url || ""
           if (retryUrl) {
@@ -212,7 +274,15 @@ export function useStep6Cover({
       clearTimeout(timeoutId)
       setGenerating(false)
     }
-  }, [selectedTemplate, assetIds, coverSettings, onCoverSettingsChange, generating, duration])
+  }, [
+    selectedTemplate,
+    assetIds,
+    coverSettings,
+    onCoverSettingsChange,
+    generating,
+    duration,
+    titleSettings,
+  ])
 
   // ── 模板操作方法 ──
   const handleSelectTemplate = useCallback((id: string) => {

apps/web/src/test/api/bgm.test.ts

@@ -32,7 +32,7 @@ describe("bgm API", () => {
 
   describe("getBgmPresets", () => {
     it("should resolve successfully", async () => {
-      await expect(getBgmPresets("test-params?")).resolves.not.toThrow()
+      await expect(getBgmPresets("test-template", { category: "test" })).resolves.not.toThrow()
     })
 
     it("should reject on API error", async () => {
@@ -42,7 +42,7 @@ describe("bgm API", () => {
       mockDelete.mockRejectedValue(new Error("Network error"))
       mockPatch.mockRejectedValue(new Error("Network error"))
 
-      await expect(getBgmPresets("test-params?")).rejects.toThrow()
+      await expect(getBgmPresets("test-template", { category: "test" })).rejects.toThrow()
     })
   })
 })

apps/web/src/test/api/edit-plans.test.ts

@@ -1,16 +1,12 @@
 import { describe, expect, it, vi, beforeEach } from "vitest"
 import {
-  getEditPlans,
   getEditPlan,
-  createEditPlan,
   updateEditPlan,
-  deleteEditPlan,
   generateEditPlan,
   getGenerationStatus,
   aiRecommendClips,
   getEditPlanGenerations,
   getGenerationTaskResults,
-  cancelGeneration,
   getEditPlanClips,
   getEditPlanClip,
   createEditPlanClip,
@@ -19,7 +15,6 @@ import {
   reorderEditPlanClips,
   batchDeleteEditPlanClips,
   createClipsFromAssets,
-  copyEditPlan,
   getMediaAssets,
   getMediaAsset,
 } from "@/api/template-editor"
@@ -53,22 +48,6 @@ describe("editPlans API", () => {
     mockPatch.mockResolvedValue({ data: { success: true, items: [] } })
   })
 
-  describe("getEditPlans", () => {
-    it("should resolve successfully", async () => {
-      await expect(getEditPlans("test-params?")).resolves.not.toThrow()
-    })
-
-    it("should reject on API error", async () => {
-      mockGet.mockRejectedValue(new Error("Network error"))
-      mockPost.mockRejectedValue(new Error("Network error"))
-      mockPut.mockRejectedValue(new Error("Network error"))
-      mockDelete.mockRejectedValue(new Error("Network error"))
-      mockPatch.mockRejectedValue(new Error("Network error"))
-
-      await expect(getEditPlans("test-params?")).rejects.toThrow()
-    })
-  })
-
   describe("getEditPlan", () => {
     it("should resolve successfully", async () => {
       await expect(getEditPlan("test-planId")).resolves.not.toThrow()
@@ -85,22 +64,6 @@ describe("editPlans API", () => {
     })
   })
 
-  describe("createEditPlan", () => {
-    it("should resolve successfully", async () => {
-      await expect(createEditPlan({ name: "test-item" })).resolves.not.toThrow()
-    })
-
-    it("should reject on API error", async () => {
-      mockGet.mockRejectedValue(new Error("Network error"))
-      mockPost.mockRejectedValue(new Error("Network error"))
-      mockPut.mockRejectedValue(new Error("Network error"))
-      mockDelete.mockRejectedValue(new Error("Network error"))
-      mockPatch.mockRejectedValue(new Error("Network error"))
-
-      await expect(createEditPlan({ name: "test-item" })).rejects.toThrow()
-    })
-  })
-
   describe("updateEditPlan", () => {
     it("should resolve successfully", async () => {
       await expect(updateEditPlan("test-planId")).resolves.not.toThrow()
@@ -117,22 +80,6 @@ describe("editPlans API", () => {
     })
   })
 
-  describe("deleteEditPlan", () => {
-    it("should resolve successfully", async () => {
-      await expect(deleteEditPlan("test-planId")).resolves.not.toThrow()
-    })
-
-    it("should reject on API error", async () => {
-      mockGet.mockRejectedValue(new Error("Network error"))
-      mockPost.mockRejectedValue(new Error("Network error"))
-      mockPut.mockRejectedValue(new Error("Network error"))
-      mockDelete.mockRejectedValue(new Error("Network error"))
-      mockPatch.mockRejectedValue(new Error("Network error"))
-
-      await expect(deleteEditPlan("test-planId")).rejects.toThrow()
-    })
-  })
-
   describe("generateEditPlan", () => {
     it("should resolve successfully", async () => {
       await expect(generateEditPlan("test-planId")).resolves.not.toThrow()
@@ -213,22 +160,6 @@ describe("editPlans API", () => {
     })
   })
 
-  describe("cancelGeneration", () => {
-    it("should resolve successfully", async () => {
-      await expect(cancelGeneration("test-planId")).resolves.not.toThrow()
-    })
-
-    it("should reject on API error", async () => {
-      mockGet.mockRejectedValue(new Error("Network error"))
-      mockPost.mockRejectedValue(new Error("Network error"))
-      mockPut.mockRejectedValue(new Error("Network error"))
-      mockDelete.mockRejectedValue(new Error("Network error"))
-      mockPatch.mockRejectedValue(new Error("Network error"))
-
-      await expect(cancelGeneration("test-planId")).rejects.toThrow()
-    })
-  })
-
   describe("getEditPlanClips", () => {
     it("should resolve successfully", async () => {
       await expect(getEditPlanClips("test-planId")).resolves.not.toThrow()
@@ -357,22 +288,6 @@ describe("editPlans API", () => {
     })
   })
 
-  describe("copyEditPlan", () => {
-    it("should resolve successfully", async () => {
-      await expect(copyEditPlan("test-planId")).resolves.not.toThrow()
-    })
-
-    it("should reject on API error", async () => {
-      mockGet.mockRejectedValue(new Error("Network error"))
-      mockPost.mockRejectedValue(new Error("Network error"))
-      mockPut.mockRejectedValue(new Error("Network error"))
-      mockDelete.mockRejectedValue(new Error("Network error"))
-      mockPatch.mockRejectedValue(new Error("Network error"))
-
-      await expect(copyEditPlan("test-planId")).rejects.toThrow()
-    })
-  })
-
   describe("getMediaAssets", () => {
     it("should resolve successfully", async () => {
       await expect(getMediaAssets("test-libraryId?")).resolves.not.toThrow()

apps/web/src/test/api/templates.test.ts

@@ -5,7 +5,6 @@ import {
   getTemplate,
   toggleFavoriteTemplate,
   copyTemplate,
-  generateFromTemplate,
 } from "@/api/templates"
 
 const mockGet = vi.fn()
@@ -116,20 +115,4 @@ describe("templates API", () => {
       await expect(copyTemplate("test-templateId")).rejects.toThrow()
     })
   })
-
-  describe("generateFromTemplate", () => {
-    it("should resolve successfully", async () => {
-      await expect(generateFromTemplate("test-templateId")).resolves.not.toThrow()
-    })
-
-    it("should reject on API error", async () => {
-      mockGet.mockRejectedValue(new Error("Network error"))
-      mockPost.mockRejectedValue(new Error("Network error"))
-      mockPut.mockRejectedValue(new Error("Network error"))
-      mockDelete.mockRejectedValue(new Error("Network error"))
-      mockPatch.mockRejectedValue(new Error("Network error"))
-
-      await expect(generateFromTemplate("test-templateId")).rejects.toThrow()
-    })
-  })
 })

apps/web/src/test/pages/EditingPlanner.test.tsx

@@ -150,12 +150,10 @@ vi.mock("@/api/template-editor", () => ({
   getMediaAssets: vi.fn().mockResolvedValue({ items: [] }),
   getEditPlanGenerations: vi.fn().mockResolvedValue({ items: [] }),
   getEditPlan: vi.fn().mockResolvedValue({}),
-  createEditPlan: vi.fn().mockResolvedValue({ id: "test-plan" }),
   updateEditPlan: vi.fn().mockResolvedValue({}),
   generateEditPlan: vi.fn().mockResolvedValue({ task_id: "test-task" }),
   getGenerationStatus: vi.fn().mockResolvedValue({ status: "completed" }),
   getGenerationTaskResults: vi.fn().mockResolvedValue({ items: [] }),
-  cancelGeneration: vi.fn().mockResolvedValue({}),
   getEditPlanClips: vi.fn().mockResolvedValue({ items: [] }),
   createEditPlanClip: vi.fn().mockResolvedValue({}),
   batchDeleteEditPlanClips: vi.fn().mockResolvedValue({}),

apps/web/src/test/pages/TemplateLibrary.test.tsx

@@ -109,7 +109,6 @@ vi.mock("@/api/templates", () => ({
   getTemplate: vi.fn().mockResolvedValue({ items: [], total: 0, success: true }),
   toggleFavoriteTemplate: vi.fn().mockResolvedValue({ items: [], total: 0, success: true }),
   copyTemplate: vi.fn().mockResolvedValue({ items: [], total: 0, success: true }),
-  generateFromTemplate: vi.fn().mockResolvedValue({ items: [], total: 0, success: true }),
 }))
 
 vi.mock("@/pages/templates/TemplateLibrary.css", () => ({}))

apps/worker/video_processing/__init__.py

@@ -7,7 +7,7 @@ VideoProcessor 等）按需从子模块导入，避免 __init__ 阶段引入
 packages / DB 等重依赖。
 """
 
-# 共享工具模块（零外部依赖，供 editing_modes / generation / edit_plan_generation 等复用）
+# 共享工具模块（零外部依赖，供 editing_modes / generation 等复用）
 from . import dedup_helpers, ffmpeg_utils, oss_helpers, url_security
 
 __all__ = [

apps/worker/video_processing/dedup_helpers.py

@@ -1,6 +1,6 @@
 """查重辅助函数 — 从 generation.py 提取的 GeneratedVideo 记录 + 查重逻辑.
 
-供 render_edit_plan 和 generate_video 共同复用，
+供 generate_video 共同复用，
 创建 GeneratedVideo 记录后计算指纹并执行项目级 + 批次内查重。
 """
 

apps/worker/video_processing/oss_helpers.py

@@ -1,4 +1,4 @@
-"""OSS 工具函数 — 从 generation.py / edit_plan_generation.py 提取的共享 OSS 操作.
+"""OSS 工具函数 — 从 generation.py 提取的共享 OSS 操作.
 
 提供 OSS 配置读取、Bucket 创建、素材上传/下载、asset_id → 本地路径解析
 等能力，供 render_edit_plan 和 generate_video 共同复用。

apps/worker/video_processing/render_adapter.py

@@ -128,6 +128,7 @@ class RenderAdapter:
         job_id: str = "",
         work_dir: Path | None = None,
         progress_cb: ProgressCallback | None = None,
+        voiceover_audio_path: str | None = None,
     ) -> RenderAdapterResult:
         """渲染一个 EditPlan。
 
@@ -142,6 +143,7 @@ class RenderAdapter:
             job_id: 关联的 Job ID（用于结果存储路径）
             work_dir: 工作目录，不传则使用临时目录
             progress_cb: 进度回调函数
+            voiceover_audio_path: 配音音频本地路径（一键生成场景使用）
 
         Returns:
             RenderAdapterResult
@@ -208,6 +210,7 @@ class RenderAdapter:
                 progress_cb=progress_cb,
                 rendered_clip_ids=rendered_clip_ids,
                 failed_clip_ids=failed_clip_ids,
+                voiceover_audio_path=voiceover_audio_path,
             )
 
         except subprocess.CalledProcessError as exc:
@@ -585,14 +588,11 @@ class RenderAdapter:
         try:
             from video_processing.thumbnail_generator import extract_and_upload_cover_frames
 
-            # 从 plan config 提取标题文字，叠加到封面候选帧上
-            _title_cfg = (plan_config or {}).get("title", {}) or {}
-            if not isinstance(_title_cfg, dict):
-                _title_cfg = {}
-            _title_text = (_title_cfg.get("text", "") or "").strip() if _title_cfg.get("enabled", True) else ""
-
+            # 已渲染视频在统一渲染阶段已通过 ASS 字幕把标题烧录进画面，
+            # 抽帧天然带标题，因此这里传空字符串，避免 Pillow 二次叠加导致重影。
+            # Pillow 叠加仅用于 API 从源素材抽帧（源素材本身无标题）的兜底场景。
             cover_candidates = extract_and_upload_cover_frames(
-                str(result.output_path), plan_id, num_frames=3, title_text=_title_text
+                str(result.output_path), plan_id, num_frames=3, title_text=""
             )
             if cover_candidates:
                 logger.info(

apps/worker/video_processing/thumbnail_generator.py

@@ -1,7 +1,8 @@
-"""视频封面抽帧工具 — 从已渲染视频中抽取帧作为封面。
+"""视频封面抽帧工具 — 从视频中抽取帧作为封面，支持标题文字叠加。
 
-统一封面管道：视频渲染时标题已通过 ASS 字幕烧进视频，
-渲染完成后直接从此视频抽帧，封面天然带标题，无需额外叠加逻辑。
+统一封面管道：
+- 从已渲染视频抽帧：标题已通过 ASS 字幕烧进视频，帧天然带标题，无需再叠加。
+- 从源素材抽帧（API E2 兜底）：源素材无标题，通过 Pillow 在帧上绘制标题文字。
 """
 
 from __future__ import annotations
@@ -12,6 +13,40 @@ from pathlib import Path
 
 logger = logging.getLogger(__name__)
 
+# ── 标题叠加（Pillow）──────────────────────────────────────────────────────
+# 实现统一放在 packages/shared/title_overlay.py，API 和 Worker 共用。
+
+
+def apply_title_overlay(
+    image_path: str,
+    title_text: str,
+    *,
+    color: str = "#ffffff",
+    position: str = "bottom",
+    font_size: int | None = None,
+    margin_ratio: float = 0.06,
+    stroke_width_ratio: float = 0.04,
+) -> str:
+    """在图片上绘制标题文字（指定颜色 + 黑色描边/阴影）。
+
+    委托给 packages.shared.title_overlay.apply_title_to_image，
+    保持 Worker 内调用方式不变。title_text 为空时直接返回原路径。
+    """
+    from packages.shared.title_overlay import apply_title_to_image
+
+    if not title_text or not title_text.strip():
+        return image_path
+    result = apply_title_to_image(
+        image_path,
+        title_text,
+        color=color,
+        position=position,
+        font_size=font_size,
+        margin_ratio=margin_ratio,
+        stroke_width_ratio=stroke_width_ratio,
+    )
+    return result or image_path
+
 
 def extract_first_frame(
     video_path: str,
@@ -175,6 +210,9 @@ def extract_and_upload_cover_frames(
     *,
     num_frames: int = 3,
     title_text: str = "",
+    title_color: str = "#ffffff",
+    title_position: str = "bottom",
+    title_font_size: int | None = None,
 ) -> list[dict]:
     """从视频中抽取多帧作为封面候选，上传到 OSS。
 
@@ -182,7 +220,11 @@ def extract_and_upload_cover_frames(
         video_path: 视频文件路径
         plan_id: 编辑计划 ID（用于生成 storage key）
         num_frames: 抽取帧数（默认 3）
-        title_text: 标题文字（当前版本未叠加，预留参数）
+        title_text: 标题文字；非空时用 Pillow 叠加到每帧。
+            从已渲染视频抽帧时通常传空（标题已烧录）；从源素材抽帧时传标题。
+        title_color: 标题字体颜色（#RRGGBB）
+        title_position: 标题位置 top/center/bottom
+        title_font_size: 标题字号，None 时自动计算
 
     Returns:
         封面候选列表，每项包含 {"url": str, "position": float}
@@ -208,6 +250,15 @@ def extract_and_upload_cover_frames(
                 seek_ratio=ratio,
                 min_seek_seconds=0.5,
             )
+            # 从源素材抽帧时叠加标题文字；已渲染视频标题已烧录时传空字符串跳过
+            if title_text and title_text.strip():
+                apply_title_overlay(
+                    frame_path,
+                    title_text,
+                    color=title_color,
+                    position=title_position,
+                    font_size=title_font_size,
+                )
             storage_key = f"covers/{plan_id}/frame_{i}.jpg"
             url = upload_to_oss(frame_path, storage_key)
             if url:

apps/worker/worker_app/celery_app.py

@@ -14,9 +14,17 @@ celery_app.conf.imports = (
     "worker_app.tasks.voice_extraction",
     "worker_app.tasks.voice_clone",
     "worker_app.tasks.tts_synthesis",
-    "worker_app.tasks.edit_plan_generation",
-    "worker_app.tasks.compose_video",
     "worker_app.tasks.batch_download",
     "worker_app.tasks._startup",
     "apps.worker.video_processing.dedup",
+    "worker_app.tasks.cleanup",
 )
+
+# Celery Beat 定时任务调度
+celery_app.conf.beat_schedule = {
+    "cleanup-stale-pending-tasks": {
+        "task": "worker.cleanup_stale_pending_tasks",
+        "schedule": 600.0,  # 每 10 分钟（秒）
+        "options": {"expires": 300},  # 5 分钟过期，避免堆积
+    },
+}

apps/worker/worker_app/tasks/__init__.py

@@ -25,10 +25,6 @@ def __getattr__(name: str):
         from .voice_extraction import extract_voice_task
 
         return extract_voice_task
-    elif name == "compose_video":
-        from .compose_video import compose_video
-
-        return compose_video
     elif name == "extract_background_task":
         from .voice_extraction import extract_background_task
 
@@ -58,7 +54,6 @@ def __getattr__(name: str):
 
 __all__ = [
     "classify_asset",
-    "compose_video",
     "generate_video",
     "healthcheck",
     "ingest_asset",

apps/worker/worker_app/tasks/_startup.py

@@ -10,6 +10,9 @@ logger = logging.getLogger(__name__)
 # 孤儿任务超时阈值：渲染任务超过此时间未更新则视为卡死
 ORPHAN_TASK_TIMEOUT_MINUTES = 10
 
+# Pending 任务超时阈值：pending 任务在队列中等待超过此时间则自动清理
+PENDING_TASK_TIMEOUT_MINUTES = 30
+
 
 def cleanup_orphan_tasks(timeout_minutes: int = ORPHAN_TASK_TIMEOUT_MINUTES) -> int:  # pragma: no cover
     """清理数据库中超时未更新的 running GenerationTask（孤儿任务）。
@@ -83,6 +86,38 @@ def cleanup_stale_jobs(timeout_minutes: int = ORPHAN_TASK_TIMEOUT_MINUTES) -> in
         return 0
 
 
+def cleanup_stale_pending_tasks(timeout_minutes: int = PENDING_TASK_TIMEOUT_MINUTES) -> int:  # pragma: no cover
+    """清理数据库中卡在 pending 状态超时的 GenerationTask。
+
+    全局任务队列有 pending 数量上限，长期卡在 pending 的任务会占满队列，
+    导致新用户无法创建任务。通过 created_at 超时判断并标记为 failed。
+
+    Args:
+        timeout_minutes: 超时时间（分钟），默认 PENDING_TASK_TIMEOUT_MINUTES
+
+    Returns:
+        清理的任务数量
+    """
+    from packages.adapters.sqlalchemy_impl.generation_task_repository import (
+        SQLAlchemyGenerationTaskRepository,
+    )
+
+    session = SessionLocal()
+    try:
+        repo = SQLAlchemyGenerationTaskRepository(session)
+        count = repo.cleanup_stale_pending(timeout_minutes)
+        if count > 0:
+            logger.warning("清理了 %d 个超时的 pending GenerationTask（超过 %d 分钟未处理）", count, timeout_minutes)
+        else:
+            logger.info("无超时 pending GenerationTask 需要清理")
+        return count
+    except Exception as e:
+        logger.error("清理超时 pending GenerationTask 失败: %s", e, exc_info=True)
+        return 0
+    finally:
+        session.close()
+
+
 def cleanup_all_stale_tasks(timeout_minutes: int = ORPHAN_TASK_TIMEOUT_MINUTES) -> dict:  # pragma: no cover
     """统一清理所有超时的孤儿任务。
 
@@ -93,15 +128,17 @@ def cleanup_all_stale_tasks(timeout_minutes: int = ORPHAN_TASK_TIMEOUT_MINUTES)
     """
     gen_count = cleanup_orphan_tasks(timeout_minutes)
     job_count = cleanup_stale_jobs(timeout_minutes)
-    total = gen_count + job_count
+    pending_count = cleanup_stale_pending_tasks(PENDING_TASK_TIMEOUT_MINUTES)
+    total = gen_count + job_count + pending_count
     if total > 0:
         logger.warning(
-            "孤儿任务清理完成: GenerationTask=%d, Job=%d, 总计=%d",
+            "任务清理完成: 孤儿 GenerationTask=%d, 孤儿 Job=%d, 超时 pending=%d, 总计=%d",
             gen_count,
             job_count,
+            pending_count,
             total,
         )
-    return {"generation_tasks": gen_count, "jobs": job_count}
+    return {"generation_tasks": gen_count, "jobs": job_count, "pending": pending_count}
 
 
 @worker_ready.connect

apps/worker/worker_app/tasks/cleanup.py

@@ -0,0 +1,35 @@
+"""定期清理任务 — Celery Beat 调度。
+
+包含：
+- cleanup_stale_pending_tasks: 定期清理卡在 pending 超时的 generation_tasks
+"""
+
+import logging
+
+from celery import shared_task
+from worker_app.tasks._startup import (
+    PENDING_TASK_TIMEOUT_MINUTES,
+    cleanup_stale_pending_tasks,
+)
+
+logger = logging.getLogger(__name__)
+
+
+@shared_task(name="worker.cleanup_stale_pending_tasks")
+def scheduled_cleanup_stale_pending(timeout_minutes: int = PENDING_TASK_TIMEOUT_MINUTES) -> dict:
+    """Celery Beat 调度的定期任务：清理超时的 pending 任务。
+
+    每 10 分钟执行一次（由 celery_app.py 的 beat_schedule 配置），
+    查找所有 status='pending' 且 created_at < NOW() - timeout_minutes
+    的 generation_tasks，批量更新为 failed。
+
+    Args:
+        timeout_minutes: 超时时间（分钟），默认 30 分钟
+
+    Returns:
+        {"cleaned": int}
+    """
+    count = cleanup_stale_pending_tasks(timeout_minutes)
+    if count > 0:
+        logger.info("[Beat] 清理了 %d 个超时 pending 任务（超时阈值 %d 分钟）", count, timeout_minutes)
+    return {"cleaned": count}

apps/worker/worker_app/tasks/compose_video.py

@@ -1,198 +0,0 @@
-"""视频合成 Celery 任务 — Phase 8 任务 2.10.
-
-使用 JobService 管理任务生命周期，通过 RenderAdapter 调用 UnifiedRenderService 执行合成。
-"""
-
-from __future__ import annotations
-
-import os
-import subprocess  # pragma: no cover
-import tempfile
-from pathlib import Path
-
-from celery.exceptions import SoftTimeLimitExceeded  # pragma: no cover
-from celery.utils.log import get_task_logger
-from worker_app.celery_app import celery_app
-from worker_app.db import SessionLocal
-
-logger = get_task_logger(__name__)
-
-# 任务超时时间（秒）：超过此时间 Celery 会抛出 SoftTimeLimitExceeded
-RENDER_TASK_SOFT_TIME_LIMIT = 600  # pragma: no cover  # 10 分钟
-# 硬超时：超过此时间进程会被强制 kill
-RENDER_TASK_TIME_LIMIT = 660  # pragma: no cover  # 10 分钟 + 1 分钟清理缓冲
-
-
-def _get_job_service():
-    """延迟导入 JobService，避免循环依赖。"""
-    from apps.api.app.services.job_service import JobService
-    from packages.adapters.sqlalchemy_impl.job_repository import SQLAlchemyJobRepository
-
-    db = SessionLocal()
-    repo = SQLAlchemyJobRepository(db)
-    return JobService(repo), db
-
-
-@celery_app.task(
-    name="worker.compose_video",
-    bind=True,
-    max_retries=3,
-    default_retry_delay=60,
-    soft_time_limit=RENDER_TASK_SOFT_TIME_LIMIT,
-    time_limit=RENDER_TASK_TIME_LIMIT,
-)
-def compose_video(self, job_id: str, **kwargs):  # pragma: no cover
-    """视频合成任务。
-
-    使用 UnifiedRenderService（图层架构）进行渲染。
-
-    Args:
-        job_id: JobService 中的任务 ID
-        **kwargs: 来自 Job.payload 的额外参数（plan_id, output_path 等）
-    """
-    job_service, db = _get_job_service()
-
-    try:
-        job = job_service.get_job(job_id)
-        if job is None:
-            logger.error("Job not found: %s", job_id)
-            return {"status": "error", "message": f"Job {job_id} not found"}
-
-        plan_id = job.payload.get("plan_id", "")
-        if not plan_id:
-            job_service.fail_job(job_id, "Missing plan_id in job payload")
-            return {"status": "error", "message": "Missing plan_id"}
-
-        # 使用 unified 渲染引擎
-        return _compose_with_unified_engine(self, job_service, job, plan_id, db)
-
-    except SoftTimeLimitExceeded:
-        # Celery 软超时：任务执行超过 soft_time_limit
-        error_msg = f"渲染任务超时（超过 {RENDER_TASK_SOFT_TIME_LIMIT // 60} 分钟）"
-        logger.error("视频合成超时: job_id=%s", job_id)
-        try:
-            job_service.fail_job(job_id, error_msg)
-        except Exception:
-            logger.exception("更新 Job 超时失败状态时出错")
-        # 超时不重试
-        return {"status": "error", "message": error_msg, "error_type": "timeout"}
-
-    except subprocess.TimeoutExpired as exc:
-        # FFmpeg 子进程超时
-        error_msg = f"FFmpeg 渲染超时（{exc.timeout}s）"
-        logger.error("视频合成 FFmpeg 超时: job_id=%s timeout=%s", job_id, exc.timeout)
-        try:
-            job_service.fail_job(job_id, error_msg)
-        except Exception:
-            logger.exception("更新 Job 超时失败状态时出错")
-        # 超时不重试
-        return {"status": "error", "message": error_msg, "error_type": "ffmpeg_timeout"}
-
-    except self.retry_exc as exc:
-        logger.warning("视频合成重试中: job_id=%s, exc=%s", job_id, exc)
-        raise
-
-    except subprocess.CalledProcessError as exc:
-        # FFmpeg 执行失败，提取有意义的错误信息
-        from video_processing.video_validation import get_exit_code_message
-
-        exit_msg = get_exit_code_message(exc.returncode)
-        stderr_text = (exc.stderr or "").strip()
-        stderr_tail = stderr_text[-300:] if len(stderr_text) > 300 else stderr_text
-        error_msg = f"渲染失败: {exit_msg}"
-        if stderr_tail:
-            error_msg += f" | {stderr_tail[:200]}"
-
-        logger.error("视频合成 FFmpeg 失败: job_id=%s %s", job_id, exit_msg)
-        try:
-            job_service.fail_job(job_id, error_msg[:500])
-        except Exception:
-            logger.exception("更新 Job 失败状态时出错")
-        # FFmpeg 错误不重试（通常是素材或配置问题）
-        return {"status": "error", "message": error_msg, "error_type": "ffmpeg_error", "exit_code": exc.returncode}
-
-    except Exception as exc:
-        logger.exception("视频合成异常: job_id=%s", job_id)
-        try:
-            job_service.fail_job(job_id, str(exc)[:500])
-        except Exception:
-            logger.exception("更新 Job 失败状态时出错")
-        raise self.retry(exc=exc, countdown=60) from exc
-    finally:
-        db.close()
-
-
-def _compose_with_unified_engine(task, job_service, job, plan_id: str, db) -> dict:  # pragma: no cover
-    """新引擎渲染路径（UnifiedRenderService + RenderAdapter）。"""
-    job_id = job.id
-
-    # 标记为 running
-    job_service.update_progress(job_id, progress=10.0, current_stage="初始化统一渲染引擎")
-
-    from video_processing.render_adapter import RenderAdapter
-
-    adapter = RenderAdapter(db)
-
-    # 校验合成条件
-    job_service.update_progress(job_id, progress=15.0, current_stage="校验合成条件")
-    valid, errors, warnings, ready_count, total_count = adapter.validate_plan(plan_id)
-    if not valid:
-        error_msg = "; ".join(errors)
-        job_service.fail_job(job_id, f"合成校验失败: {error_msg}")
-        return {"status": "error", "message": error_msg}
-
-    # 进度回调
-    def progress_cb(progress: float, stage: str) -> None:
-        try:
-            job_service.update_progress(job_id, progress=progress, current_stage=stage)
-        except Exception:
-            logger.exception("更新进度失败")
-
-    # 执行渲染
-    job_service.update_progress(job_id, progress=20.0, current_stage="开始渲染")
-    logger.info("统一渲染引擎开始: job_id=%s plan_id=%s", job_id, plan_id)
-
-    result = adapter.render_plan(
-        plan_id=plan_id,
-        job_id=job_id,
-        progress_cb=progress_cb,
-    )
-
-    if not result.success:
-        error_msg = f"渲染失败: {result.error_message}"
-        job_service.fail_job(job_id, error_msg[:500])
-        raise RuntimeError(result.error_message)
-
-    # 更新 Job 状态为完成
-    result_data = {
-        "plan_id": plan_id,
-        "output_path": str(result.output_path) if result.output_path else "",
-        "storage_key": f"rendered/{plan_id}/{job_id}.mp4",
-        "output_url": result.output_url,
-        "estimated_duration": result.duration,
-        "clip_count": result.clip_count,
-        "engine": "unified",
-        "width": result.width,
-        "height": result.height,
-        "file_size": result.file_size,
-    }
-    job_service.complete_job(job_id, result=result_data)
-
-    logger.info(
-        "视频合成完成(unified): job_id=%s plan_id=%s duration=%.2fs",
-        job_id,
-        plan_id,
-        result.duration,
-    )
-    return {"status": "completed", "job_id": job_id, "result": result_data}
-
-
-def _cleanup_output(job_id: str) -> None:
-    """清理临时输出文件。"""
-    try:
-        _output_dir = os.environ.get("VIDEO_OUTPUT_DIR", os.path.join(tempfile.gettempdir(), "video_output"))
-        output_path = os.path.join(_output_dir, f"{job_id}.mp4")
-        if Path(output_path).exists():
-            Path(output_path).unlink()
-    except Exception as e:
-        logger.warning(f"清理输出文件失败: {e}", exc_info=True)

apps/worker/worker_app/tasks/edit_plan_generation.py

@@ -1,452 +0,0 @@
-"""剪辑计划渲染任务 — 使用 UnifiedRenderService 统一渲染引擎.
-
-Celery 任务 worker.render_edit_plan:
-  1. 加载 EditPlan + EditPlanClips
-  2. 通过 RenderAdapter 调用 UnifiedRenderService 渲染
-  3. 下载各片段素材 + 渲染
-  4. 上传渲染结果到 OSS
-  5. 创建 GeneratedVideo 记录 + 查重
-  6. 更新 EditPlan / EditPlanClip 状态
-  7. 更新 GenerationTask 进度
-"""
-
-from __future__ import annotations
-
-import logging
-from datetime import datetime, timezone
-from pathlib import Path
-
-from worker_app.celery_app import celery_app
-from worker_app.db import SessionLocal
-
-logger = logging.getLogger(__name__)
-
-OUTPUT_WIDTH = 1280
-OUTPUT_HEIGHT = 720
-OUTPUT_FPS = 25.0
-
-
-# ── 共享工具模块导入 ──────────────────────────────────────────────────────────
-
-from video_processing.dedup_helpers import create_video_record_and_dedup
-
-# ── Repository imports (延迟导入避免循环依赖) ─────────────────────────────────
-
-
-def _get_repos():
-    """获取数据库仓储实例"""
-    from packages.adapters.sqlalchemy_impl import SQLAlchemyEditPlanRepository
-    from packages.adapters.sqlalchemy_impl.edit_plan_clip_repository import (
-        SQLAlchemyEditPlanClipRepository,
-    )
-    from packages.adapters.sqlalchemy_impl.generation_task_repository import (
-        SQLAlchemyGenerationTaskRepository,
-    )
-
-    db = SessionLocal()
-    try:
-        plan_repo = SQLAlchemyEditPlanRepository(db)
-        clip_repo = SQLAlchemyEditPlanClipRepository(db)
-        gen_task_repo = SQLAlchemyGenerationTaskRepository(db)
-        yield plan_repo, clip_repo, gen_task_repo, db
-    finally:
-        db.close()
-
-
-# ── Celery Task ───────────────────────────────────────────────────────────────
-
-
-def _mark_plan_failed(plan_repo, plan_id, gen_task_repo, generation_task_id, error_msg: str):
-    """统一的计划失败标记工具。"""
-    plan = plan_repo.get(plan_id)
-    if plan and plan.status.value == "rendering":
-        plan.mark_failed()
-        plan_repo.update(plan)
-    if generation_task_id:
-        gen_task = gen_task_repo.get(generation_task_id)
-        if gen_task and gen_task.status.value != "failed":
-            gen_task.status = "failed"
-            gen_task.error_message = error_msg
-            gen_task.completed_at = datetime.now(timezone.utc)
-            try:
-                gen_task.append_log(
-                    stage="render_failed",
-                    message=error_msg[:500],
-                    level="ERROR",
-                )
-            except Exception:
-                pass
-            gen_task_repo.update(gen_task)
-
-
-def _finalize_render_success(
-    plan,
-    plan_repo,
-    clip_repo,
-    gen_task_repo,
-    db,
-    plan_id: str,
-    output_url: str,
-    storage_key: str,
-    duration: float,
-    file_size: int,
-    width: int,
-    height: int,
-    rendered_clip_ids: list[str],
-    failed_clip_ids: list[str],
-    generation_task_id: str,
-    output_path: Path,
-    engine: str,
-    thumbnail_url: str = "",
-    cover_candidates: list[dict] | None = None,
-) -> dict:
-    """渲染成功后的统一收尾：查重 + 更新状态 + 返回结果。"""
-    # 创建 GeneratedVideo 记录 + 查重
-    project_id = plan.project_id or ""
-    batch_id = plan.config.get("batch_id", "")
-    mode = plan.config.get("mode", "edit_plan")
-    # 从 plan.config.title.text 读取视频名称
-    plan_config = plan.config or {}
-    title_cfg = plan_config.get("title", {}) or {}
-    if not isinstance(title_cfg, dict):
-        title_cfg = {}
-    video_name = (title_cfg.get("text") or "").strip() or f"generated-{generation_task_id[:8]}.mp4"
-    if generation_task_id:
-        try:
-            create_video_record_and_dedup(
-                generation_task_id=generation_task_id,
-                project_id=project_id,
-                user_id=plan.created_by_user_id or "",
-                batch_id=batch_id,
-                file_url=output_url or "",
-                file_size=file_size,
-                duration=duration,
-                video_path=str(output_path),
-                mode=mode,
-                session=db,
-                width=width,
-                height=height,
-                fps=OUTPUT_FPS,
-                name=video_name,
-                thumbnail_url=thumbnail_url,
-            )
-        except Exception as dedup_err:
-            logger.warning("查重失败（不影响渲染结果）: %s", dedup_err)
-
-    # 更新片段状态为 rendered
-    for clip_id in rendered_clip_ids:
-        clip = clip_repo.get(clip_id)
-        if clip and clip.status.value == "ready":
-            clip.mark_rendered()
-            clip_repo.update(clip)
-
-    # 更新 EditPlan 状态为 completed + 回写实际渲染时长 + 结果数
-    plan.config["rendered_url"] = output_url or ""
-    plan.config["rendered_storage_key"] = storage_key
-    if hasattr(plan, "total_duration") and duration > 0:
-        plan.total_duration = duration
-    plan.mark_completed()
-    plan_repo.update(plan)
-
-    # 更新 GenerationTask 状态为 completed
-    if generation_task_id:
-        gen_task = gen_task_repo.get(generation_task_id)
-        if gen_task:
-            gen_task.status = "completed"
-            gen_task.progress = 100.0
-            # 剪辑计划是多片段合成 1 个成片，result_count = 1
-            gen_task.result_count = 1
-            gen_task.append_log(
-                stage="render_complete",
-                message=f"渲染完成，输出时长 {duration:.1f}s",
-                level="INFO",
-                engine=engine,
-                clip_count=len(rendered_clip_ids),
-            )
-            gen_task.completed_at = datetime.now(timezone.utc)
-
-            # 回写封面 URL 到 GenerationTask，供封面生成接口读取
-            if cover_candidates:
-                first_cover = cover_candidates[0].get("image_url") or cover_candidates[0].get("url") or ""
-                if first_cover:
-                    gen_task.cover_url = first_cover
-                    logger.info(
-                        "预览渲染完成，回写 cover_url: plan_id=%s task_id=%s url=%s",
-                        plan_id,
-                        generation_task_id,
-                        first_cover[:80],
-                    )
-
-            gen_task_repo.update(gen_task)
-
-    logger.info(
-        "剪辑计划渲染完成: plan_id=%s engine=%s rendered=%d failed=%d duration=%.1fs",
-        plan_id,
-        engine,
-        len(rendered_clip_ids),
-        len(failed_clip_ids),
-        duration,
-    )
-
-    return {
-        "status": "completed",
-        "plan_id": plan_id,
-        "rendered_count": len(rendered_clip_ids),
-        "failed_count": len(failed_clip_ids),
-        "output_url": output_url,
-        "duration": duration,
-    }
-
-
-def _render_with_unified(
-    plan,
-    clips,
-    plan_id: str,
-    generation_task_id: str,
-    plan_repo,
-    clip_repo,
-    gen_task_repo,
-    db,
-) -> dict:
-    """统一渲染引擎路径（通过 RenderAdapter 调用 UnifiedRenderService）。
-
-    RenderAdapter 内部处理：素材下载、BGM 准备、ASR 自动字幕、渲染执行、OSS 上传。
-    本函数只负责：业务状态更新、查重、收尾。
-    """
-    from video_processing.render_adapter import RenderAdapter
-
-    adapter = RenderAdapter(db)
-
-    # 进度回调：更新 GenerationTask 进度
-    def _progress_cb(progress: float, stage: str):
-        if not generation_task_id:
-            return
-        try:
-            gen_task = gen_task_repo.get(generation_task_id)
-            if gen_task:
-                # 映射到 30%~90% 区间（素材下载前已到 30%）
-                mapped_progress = 30.0 + progress * 0.6
-                gen_task.progress = min(mapped_progress, 95.0)
-                gen_task.append_log(
-                    stage="render_progress",
-                    message=stage,
-                    level="INFO",
-                    progress=mapped_progress,
-                )
-                gen_task_repo.update(gen_task)
-        except Exception:
-            pass
-
-    try:
-        result = adapter.render_plan(
-            plan_id=plan_id,
-            job_id=generation_task_id or plan_id,
-            progress_cb=_progress_cb,
-        )
-    except Exception as render_err:
-        logger.error("渲染失败(unified): %s — %s", plan_id, render_err)
-        _mark_plan_failed(plan_repo, plan_id, gen_task_repo, generation_task_id, f"渲染失败: {render_err}")
-        return {"status": "error", "message": f"渲染失败: {render_err}"}
-
-    if not result.success:
-        full_error = result.error_message or "渲染失败"
-        if result.error_detail:
-            full_error = f"{full_error}\n--- stderr ---\n{result.error_detail}"
-        logger.error("渲染失败(unified): %s — %s", plan_id, result.error_message)
-        _mark_plan_failed(plan_repo, plan_id, gen_task_repo, generation_task_id, full_error)
-        return {"status": "error", "message": result.error_message or "渲染失败"}
-
-    output_path = result.output_path or Path("")
-    output_url = result.output_url or ""
-    thumbnail_url = result.thumbnail_url or ""
-    # adapter 上传到 rendered/{plan_id}/{job_id}.mp4，从 URL 提取实际 key
-    # 不能用 output.mp4 硬编码，否则 cover 等下游通过 key 构造的 URL 指向不存在的文件
-    if output_url:
-        from packages.shared.storage import get_shared_storage_service
-
-        storage_key = get_shared_storage_service().normalize_storage_key(output_url)
-    else:
-        storage_key = f"rendered/{plan_id}/{generation_task_id or plan_id}.mp4"
-
-    # 用 adapter 返回的 clip 明细（以 adapter 的结果为准）
-    rendered_clip_ids = result.rendered_clip_ids or []
-    failed_clip_ids = result.failed_clip_ids or []
-
-    # 将封面候选帧写入 plan.config（供封面 API 直接使用，跳过 MediaKit 抽帧）
-    if result.cover_candidates:
-        plan_config = plan.config or {}
-        plan_config["cover_candidates"] = result.cover_candidates
-        plan.config = plan_config
-        logger.info(
-            "封面候选帧已写入 plan.config: plan_id=%s count=%d",
-            plan_id,
-            len(result.cover_candidates),
-        )
-
-    return _finalize_render_success(
-        plan=plan,
-        plan_repo=plan_repo,
-        clip_repo=clip_repo,
-        gen_task_repo=gen_task_repo,
-        db=db,
-        plan_id=plan_id,
-        output_url=output_url or "",
-        storage_key=storage_key,
-        duration=result.duration,
-        file_size=result.file_size,
-        width=result.width,
-        height=result.height,
-        rendered_clip_ids=rendered_clip_ids,
-        failed_clip_ids=failed_clip_ids,
-        generation_task_id=generation_task_id,
-        output_path=output_path,
-        engine="unified",
-        thumbnail_url=thumbnail_url,
-        cover_candidates=result.cover_candidates,
-    )
-
-
-@celery_app.task(
-    name="worker.render_edit_plan",
-    bind=True,
-    max_retries=2,
-    soft_time_limit=600,  # 10 分钟软超时
-    time_limit=660,  # 11 分钟硬超时
-)
-def render_edit_plan(self, plan_id: str) -> dict:  # pragma: no cover
-    """渲染剪辑计划
-
-    流程：
-    1. 加载 EditPlan + EditPlanClips
-    2. 通过 RenderAdapter 调用 UnifiedRenderService 渲染
-    3. 下载素材 + 渲染
-    4. 上传渲染结果到 OSS
-    5. 创建 GeneratedVideo 记录 + 查重
-    6. 更新 EditPlan → completed, EditPlanClips → rendered
-    7. 更新 GenerationTask 进度
-    """
-    logger.info("开始渲染剪辑计划: plan_id=%s", plan_id)
-
-    generation_task_id = ""
-
-    for repos in _get_repos():
-        plan_repo, clip_repo, gen_task_repo, db = repos
-
-        try:
-            # 1. 加载 EditPlan
-            plan = plan_repo.get(plan_id)
-            if plan is None:
-                logger.error("剪辑计划不存在: %s", plan_id)
-                return {"status": "error", "message": f"计划不存在: {plan_id}"}
-
-            # 获取 generation_task_id（提前读取，确保 except 块可用）
-            generation_task_id = plan.config.get("generation_task_id", "")
-
-            # 2. 准备渲染（使用 unified 渲染引擎）
-
-            # 3. 加载片段列表（按 order 排序）
-            clips = clip_repo.list_by_plan(plan_id, skip=0, limit=10000)
-            if not clips:
-                logger.warning("剪辑计划没有片段: %s", plan_id)
-                plan.mark_failed()
-                plan_repo.update(plan)
-                return {"status": "error", "message": "没有可渲染的片段"}
-
-            # 更新 GenerationTask 状态为 running
-            if generation_task_id:
-                gen_task = gen_task_repo.get(generation_task_id)
-                if gen_task:
-                    gen_task.status = "running"
-                    gen_task.started_at = datetime.now(timezone.utc)
-                    gen_task.append_log(
-                        stage="render_start",
-                        message=f"开始渲染，片段数 {len(clips)}",
-                        level="INFO",
-                        engine="unified",
-                        clip_count=len(clips),
-                    )
-                    gen_task_repo.update(gen_task)
-
-            # 3. 渲染前取消检查
-            if generation_task_id:
-                current_task = gen_task_repo.get(generation_task_id)
-                if current_task:
-                    task_status = (
-                        current_task.status.value if hasattr(current_task.status, "value") else str(current_task.status)
-                    )
-                    if task_status == "cancelled":
-                        logger.info("任务已被取消，中止渲染: plan_id=%s task_id=%s", plan_id, generation_task_id)
-
-                        if plan.status.value == "rendering":
-                            try:
-                                plan.resume_editing()
-                                plan_repo.update(plan)
-                            except ValueError:
-                                pass
-                        return {"status": "cancelled", "plan_id": plan_id, "message": "任务已取消"}
-
-            # 4. 渲染（unified 引擎：RenderAdapter 统一处理下载 + BGM + ASR + 渲染 + 上传）
-            result = _render_with_unified(
-                plan=plan,
-                clips=clips,
-                plan_id=plan_id,
-                generation_task_id=generation_task_id,
-                plan_repo=plan_repo,
-                clip_repo=clip_repo,
-                gen_task_repo=gen_task_repo,
-                db=db,
-            )
-
-            result["engine"] = "unified"
-            return result
-
-        except Exception as exc:
-            # 超时异常不重试，直接标记失败
-            from celery.exceptions import SoftTimeLimitExceeded
-
-            is_timeout = isinstance(exc, SoftTimeLimitExceeded)
-            if is_timeout:
-                logger.error("渲染剪辑计划超时: plan_id=%s", plan_id)
-            else:
-                logger.exception("渲染剪辑计划异常: %s", plan_id)
-
-            # 尝试标记计划和 GenerationTask 为失败
-            error_msg = "渲染任务超时（超过10分钟）" if is_timeout else f"渲染异常: {type(exc).__name__}: {exc}"
-            try:
-                plan = plan_repo.get(plan_id)
-                if plan and plan.status.value == "rendering":
-                    plan.mark_failed()
-                    plan_repo.update(plan)
-            except Exception as e:
-                logger.warning("标记计划失败时异常: plan_id=%s error=%s", plan_id, e, exc_info=True)
-            # 更新 GenerationTask 状态为 failed，前端轮询能看到失败状态
-            try:
-                if generation_task_id:
-                    gen_task = gen_task_repo.get(generation_task_id)
-                    if gen_task and gen_task.status.value != "failed":
-                        gen_task.status = "failed"
-                        gen_task.error_message = error_msg
-                        gen_task.completed_at = datetime.now(timezone.utc)
-                        try:
-                            log_stage = "render_timeout" if is_timeout else "render_failed"
-                            gen_task.append_log(
-                                stage=log_stage,
-                                message=error_msg[:500],
-                                level="ERROR",
-                                exception_type=type(exc).__name__,
-                            )
-                        except Exception:
-                            pass
-                        gen_task_repo.update(gen_task)
-                        logger.info(
-                            "GenerationTask 已标记为 failed: task_id=%s plan_id=%s",
-                            generation_task_id,
-                            plan_id,
-                        )
-            except Exception as e:
-                logger.warning(
-                    "更新 GenerationTask 失败状态时异常: task_id=%s error=%s", generation_task_id, e, exc_info=True
-                )
-            raise self.retry(exc=exc, countdown=60) from exc
-
-    return {"status": "error", "message": "数据库连接失败"}

apps/worker/worker_app/tasks/generation.py

@@ -223,6 +223,7 @@ def _load_template_segment_durations(template_id: str) -> list[float]:
         return []
 
 
+# DEPRECATED: 仅兼容无 source_edit_plan_id 的旧调用，后续移除
 def _build_plan_and_clips_from_task(
     task_id: str,
     downloaded_paths: list[Path],
@@ -887,7 +888,9 @@ def _load_task_info(task_id: str) -> dict | None:
             "output_height": getattr(gen_task, "output_height", OUTPUT_HEIGHT) or OUTPUT_HEIGHT,
             "cover_url": getattr(gen_task, "cover_url", "") or "",
             "custom_title": getattr(gen_task, "custom_title", "") or "",
+            "title_config": dict(getattr(gen_task, "title_config", {}) or {}),
             "voice_ids": list(getattr(gen_task, "voice_ids", []) or []),
+            "source_edit_plan_id": getattr(gen_task, "source_edit_plan_id", "") or "",
         }
     finally:
         session.close()
@@ -967,14 +970,15 @@ def _render_video(
     bgm_config: dict | None = None,
     voice_ids: list[str] | None = None,
     custom_title: str = "",
-) -> tuple[Path, float]:
+    title_config: dict | None = None,
+) -> tuple[Path, float, list[dict] | None]:
     """渲染视频（含配音混音）。
 
     使用 RenderAdapter 统一渲染入口，复用 BGM/ASR/分辨率/缩略图逻辑。
 
     Args:
     Returns:
-        (output_path, render_duration)
+        (output_path, render_duration, cover_candidates)
     """
     if not downloaded_videos:
         raise RuntimeError(f"素材下载结果为空: task_id={task_id}")
@@ -999,27 +1003,34 @@ def _render_video(
                 list(template_config.keys()),
             )
 
-    # ── 用户自定义标题覆盖模板标题配置 ──────────────────────────────────
-    if custom_title:
+    # ── 用户自定义标题：title_config 优先，custom_title 兜底 ─────────────
+    effective_title_cfg: dict | None = None
+    if title_config and isinstance(title_config, dict) and title_config.get("text", "").strip():
+        effective_title_cfg = dict(title_config)
+    elif custom_title:
         try:
-            user_title_cfg = json.loads(custom_title) if isinstance(custom_title, str) else custom_title
-            if isinstance(user_title_cfg, dict) and user_title_cfg.get("text", "").strip():
-                # 字段名归一化: 前端 font_size/font_color → 后端 size/color
-                if "font_size" in user_title_cfg and "size" not in user_title_cfg:
-                    user_title_cfg["size"] = user_title_cfg["font_size"]
-                if "font_color" in user_title_cfg and "color" not in user_title_cfg:
-                    user_title_cfg["color"] = user_title_cfg["font_color"]
-                plan_cfg = dict(virtual_plan.config or {})
-                plan_cfg["title"] = user_title_cfg
-                virtual_plan.config = plan_cfg
-                logger.info(
-                    "[task_id=%s] [渲染] 用户自定义标题已注入: text=%s",
-                    task_id,
-                    user_title_cfg.get("text", "")[:30],
-                )
+            parsed = json.loads(custom_title) if isinstance(custom_title, str) else custom_title
+            if isinstance(parsed, dict) and parsed.get("text", "").strip():
+                effective_title_cfg = parsed
         except (json.JSONDecodeError, TypeError):
             logger.warning("[task_id=%s] custom_title JSON解析失败: %s", task_id, custom_title[:100])
 
+    if effective_title_cfg:
+        # 字段名归一化: 前端 font_size/font_color → 后端 size/color
+        if "font_size" in effective_title_cfg and "size" not in effective_title_cfg:
+            effective_title_cfg["size"] = effective_title_cfg["font_size"]
+        if "font_color" in effective_title_cfg and "color" not in effective_title_cfg:
+            effective_title_cfg["color"] = effective_title_cfg["font_color"]
+        plan_cfg = dict(virtual_plan.config or {})
+        plan_cfg["title"] = effective_title_cfg
+        virtual_plan.config = plan_cfg
+        logger.info(
+            "[task_id=%s] [渲染] 标题配置已注入(source=%s): text=%s",
+            task_id,
+            "title_config" if title_config else "custom_title",
+            effective_title_cfg.get("text", "")[:30],
+        )
+
     # 用户自定义 BGM 覆盖模板 BGM（用户指定优先级最高）
     if bgm_config:
         plan_cfg = virtual_plan.config or {}
@@ -1113,8 +1124,10 @@ def _render_video(
 
     # 配音素材库音频已在统一渲染引擎内部通过 audio 图层混音处理
     output_path = render_output_path
+    # RenderAdapter 在渲染完成后用本地 ffmpeg 抽取的封面候选帧（已上传 OSS）
+    cover_candidates = getattr(render_result, "cover_candidates", None)
 
-    return output_path, render_duration
+    return output_path, render_duration, cover_candidates
 
 
 def _upload_and_record(
@@ -1200,6 +1213,135 @@ def _upload_and_record(
     soft_time_limit=600,  # 10 分钟软超时
     time_limit=660,  # 11 分钟硬超时
 )
+def _sync_task_config_to_plan(source_edit_plan_id: str, task_info: dict, db) -> str | None:
+    """将 GenerationTask 的配置同步到 EditPlan.config，返回配音本地路径（如果有）。
+
+    包括：title_config、BGM、输出分辨率。配音单独处理（需下载到本地）。
+    """
+    from packages.adapters.sqlalchemy_impl.edit_plan_repository import (
+        SQLAlchemyEditPlanRepository,
+    )
+
+    plan_repo = SQLAlchemyEditPlanRepository(db)
+    plan = plan_repo.get(source_edit_plan_id)
+    if plan is None:
+        logger.error("[task] EditPlan not found: %s", source_edit_plan_id)
+        return None
+
+    plan_config = dict(plan.config or {})
+    changed = False
+
+    # 标题配置
+    title_config = task_info.get("title_config") or {}
+    if title_config and isinstance(title_config, dict) and title_config.get("text", "").strip():
+        cfg = dict(title_config)
+        # 字段名归一化
+        if "font_size" in cfg and "size" not in cfg:
+            cfg["size"] = cfg["font_size"]
+        if "font_color" in cfg and "color" not in cfg:
+            cfg["color"] = cfg["font_color"]
+        plan_config["title"] = cfg
+        changed = True
+        logger.info("[task] title_config synced to plan: %s", cfg.get("text", "")[:30])
+
+    # BGM 配置
+    bgm_config = task_info.get("bgm_config") or {}
+    if bgm_config:
+        from packages.domain.bgm_utils import merge_bgm_config
+
+        existing_bgm = plan_config.get("bgm", {}) or {}
+        plan_config["bgm"] = merge_bgm_config(existing_bgm, bgm_config)
+        changed = True
+
+    # 输出分辨率
+    ow = task_info.get("output_width") or OUTPUT_WIDTH
+    oh = task_info.get("output_height") or OUTPUT_HEIGHT
+    if ow >= 100 and oh >= 100:
+        export_cfg = dict(plan_config.get("export", {}) or {})
+        export_cfg["resolution"] = f"{ow}x{oh}"
+        plan_config["export"] = export_cfg
+        changed = True
+
+    if changed:
+        plan.config = plan_config
+        plan_repo.update(plan)
+        logger.info("[task] plan.config synced: plan_id=%s", source_edit_plan_id)
+
+    # 配音下载
+    voiceover_path: str | None = None
+    voice_library_id = task_info.get("voice_library_id", "")
+    voice_ids = task_info.get("voice_ids", []) or []
+    effective_voice_id = voice_library_id or (voice_ids[0] if voice_ids else "")
+
+    if effective_voice_id:
+        import tempfile
+
+        voice_tmp = Path(tempfile.gettempdir()) / f"voice_{source_edit_plan_id}_{id(task_info)}.mp3"
+        try:
+            if _download_voice_asset(effective_voice_id, voice_tmp):
+                voiceover_path = str(voice_tmp)
+                logger.info("[task] voice downloaded: %s -> %s", effective_voice_id, voiceover_path)
+        except Exception:
+            logger.warning("[task] voice download failed: %s", effective_voice_id, exc_info=True)
+
+    return voiceover_path
+
+
+def _render_from_edit_plan(
+    task_id: str,
+    source_edit_plan_id: str,
+    task_info: dict,
+) -> tuple[Path, float, list[dict] | None]:
+    """从 EditPlan 数据库记录直接渲染（不再内存重建clips）。
+
+    Returns:
+        (output_path, render_duration, cover_candidates)
+    """
+    from video_processing.render_adapter import RenderAdapter
+    from worker_app.db import SessionLocal
+
+    db = SessionLocal()
+    try:
+        # 同步配置到 plan.config + 下载配音
+        voiceover_path = _sync_task_config_to_plan(source_edit_plan_id, task_info, db)
+
+        # 进度回调
+        def _progress_cb(progress: float, stage: str):
+            mapped = 40.0 + progress * 0.4
+            _update_task_progress(task_id, min(mapped, 80.0), stage)
+
+        adapter = RenderAdapter(db)
+        render_start = time.monotonic()
+        logger.info("[task_id=%s] [渲染] RenderAdapter.render_plan 开始 (plan_id=%s)", task_id, source_edit_plan_id)
+
+        result = adapter.render_plan(
+            plan_id=source_edit_plan_id,
+            job_id=task_id,
+            progress_cb=_progress_cb,
+            voiceover_audio_path=voiceover_path,
+        )
+
+        if not result.success:
+            raise RuntimeError(f"渲染失败: {result.error_message}")
+
+        render_elapsed = time.monotonic() - render_start
+        logger.info(
+            "[task_id=%s] [渲染] RenderAdapter.render_plan 完成: 耗时=%.1fs, 时长=%.2fs",
+            task_id,
+            render_elapsed,
+            result.duration,
+        )
+
+        output_path = result.output_path
+        cover_candidates = getattr(result, "cover_candidates", None)
+
+        return output_path, result.duration, cover_candidates
+    finally:
+        db.close()
+        # 清理临时配音文件
+        # voiceover_path 在外部作用域，这里不直接引用
+
+
 def generate_video(self, task_id: str) -> dict:
     """生成视频任务 — 使用 UnifiedRenderService 统一渲染。
 
@@ -1283,6 +1425,178 @@ def generate_video(self, task_id: str) -> dict:
         if template_id:
             _validate_template_exists(template_id)
 
+        # ── 新路径：有 source_edit_plan_id 时直接从数据库 EditPlan 渲染 ──
+        source_edit_plan_id = task_info.get("source_edit_plan_id", "")
+        if source_edit_plan_id:
+            logger.info(
+                "[task_id=%s] 使用 EditPlan 数据库路径渲染: plan_id=%s",
+                task_id,
+                source_edit_plan_id,
+            )
+            _update_task_progress(task_id, 30, "加载草稿数据")
+
+            if gen_task:
+                gen_task.append_log("渲染模式", "从草稿数据渲染（与预览一致）")
+                _flush_logs(task_id, gen_task)
+
+            output_path, render_duration, cover_candidates = _render_from_edit_plan(
+                task_id=task_id,
+                source_edit_plan_id=source_edit_plan_id,
+                task_info=task_info,
+            )
+
+            if gen_task:
+                gen_task.append_log("渲染", f"渲染完成, 时长={render_duration:.1f}s")
+                _flush_logs(task_id, gen_task)
+
+            _update_task_progress(task_id, 80, "渲染完成")
+
+            # ── 4. 上传 OSS + 查重记录 ───────────────────────────────
+            _update_task_progress(task_id, 85, "开始上传")
+            file_url, duration, file_size, video_count = _upload_and_record(
+                task_id=task_id,
+                output_path=output_path,
+                project_id=project_id,
+                batch_id=batch_id,
+                editing_mode=editing_mode,
+                user_id=user_id,
+                video_name=task_info.get("video_title", ""),
+            )
+
+            if gen_task:
+                gen_task.append_log(
+                    "OSS上传",
+                    f"上传成功, 大小={file_size}",
+                    file_size=file_size,
+                    file_url=file_url,
+                )
+                _flush_logs(task_id, gen_task)
+
+            _update_task_progress(task_id, 95, "上传完成")
+
+            # ── 4.5 封面帧持久化 ────────────────────────────────────────────
+            try:
+                if cover_candidates:
+                    first = cover_candidates[0]
+                    cover_frame_url = first.get("image_url") or first.get("url") or ""
+                    if cover_frame_url:
+                        _cover_session = SessionLocal()
+                        try:
+                            from packages.adapters.sqlalchemy_impl.models import (
+                                GenerationTaskModel,
+                            )
+
+                            _cover_model = (
+                                _cover_session.query(GenerationTaskModel)
+                                .filter(GenerationTaskModel.id == task_id)
+                                .first()
+                            )
+                            if _cover_model:
+                                _cover_model.cover_url = cover_frame_url
+                                meta = dict(_cover_model.metadata or {})
+                                meta["cover_candidates"] = cover_candidates
+                                _cover_model.metadata = meta
+                                _cover_session.commit()
+                        finally:
+                            _cover_session.close()
+            except Exception:
+                logger.warning("[task_id=%s] 封面帧持久化失败", task_id, exc_info=True)
+
+            # ── 5. 标记完成 ──────────────────────────────────────────────────
+            _update_task_status(task_id, "mark_completed", result_count=video_count)
+
+            # 5.1 更新标题使用次数
+            try:
+                _title_session = SessionLocal()
+                try:
+                    from packages.adapters.sqlalchemy_impl.generation_task_repository import (
+                        SQLAlchemyGenerationTaskRepository,
+                    )
+                    from packages.adapters.sqlalchemy_impl.title_library_repository import (
+                        SQLAlchemyTitleLibraryRepository,
+                    )
+
+                    _task_repo = SQLAlchemyGenerationTaskRepository(_title_session)
+                    _gen_task = _task_repo.get(task_id)
+                    if _gen_task and _gen_task.title_ids and _gen_task.created_by_user_id:
+                        _title_repo = SQLAlchemyTitleLibraryRepository(_title_session)
+                        for _tid in _gen_task.title_ids:
+                            try:
+                                _title_repo.increment_usage_count(_tid, _gen_task.created_by_user_id)
+                            except Exception:
+                                logger.warning(
+                                    "[task_id=%s] 更新标题使用次数失败: title_id=%s",
+                                    task_id,
+                                    _tid,
+                                    exc_info=True,
+                                )
+                finally:
+                    _title_session.close()
+            except Exception:
+                logger.warning("[task_id=%s] 更新标题使用次数异常", task_id, exc_info=True)
+
+            # 5.2 更新素材使用次数
+            try:
+                from worker_app.core.asset_usage import mark_asset_used_for_generation
+
+                _asset_session = SessionLocal()
+                try:
+                    from packages.adapters.sqlalchemy_impl.asset_repository import (
+                        SQLAlchemyAssetRepository,
+                    )
+                    from packages.adapters.sqlalchemy_impl.generation_task_repository import (
+                        SQLAlchemyGenerationTaskRepository,
+                    )
+
+                    _task_repo = SQLAlchemyGenerationTaskRepository(_asset_session)
+                    _asset_repo = SQLAlchemyAssetRepository(_asset_session)
+                    _gen_task = _task_repo.get(task_id)
+                    if _gen_task and _gen_task.asset_ids:
+                        for _aid in _gen_task.asset_ids:
+                            try:
+                                _asset = _asset_repo.get(_aid)
+                                if _asset:
+                                    mark_asset_used_for_generation(_asset)
+                                    _asset_repo.update(_asset)
+                            except Exception:
+                                logger.warning(
+                                    "[task_id=%s] 更新素材使用次数失败: asset_id=%s",
+                                    task_id,
+                                    _aid,
+                                    exc_info=True,
+                                )
+                finally:
+                    _asset_session.close()
+            except Exception:
+                logger.warning("[task_id=%s] 更新素材使用次数异常", task_id, exc_info=True)
+
+            if gen_task:
+                gen_task.append_log(
+                    "任务完成",
+                    f"视频生成完成: 时长={duration:.2f}s, 大小={file_size}",
+                    duration=round(duration, 2),
+                    file_size=file_size,
+                    video_count=video_count,
+                )
+                _flush_logs(task_id, gen_task)
+
+            logger.info(
+                "[task_id=%s] [任务完成] duration=%.2fs file_size=%d (edit_plan path)",
+                task_id,
+                duration,
+                file_size,
+            )
+
+            return {
+                "status": "completed",
+                "task_id": task_id,
+                "output_path": str(output_path),
+                "file_size": file_size,
+                "duration": duration,
+                "mode": editing_mode.value,
+            }
+
+        # DEPRECATED: 以下为旧路径，仅兼容无 source_edit_plan_id 的旧调用，后续移除
         with tempfile.TemporaryDirectory(prefix="xiaoxia-generation-") as temp_dir:
             temp_path = Path(temp_dir)
 
@@ -1385,7 +1699,7 @@ def generate_video(self, task_id: str) -> dict:
             else:
                 _resolved_resolution = task_info.get("resolution", "")
 
-            output_path, render_duration = _render_video(
+            output_path, render_duration, cover_candidates = _render_video(
                 task_id=task_id,
                 downloaded_videos=downloaded_videos,
                 voice_path=audio_path,
@@ -1399,6 +1713,7 @@ def generate_video(self, task_id: str) -> dict:
                 bgm_config=task_info.get("bgm_config", {}),
                 voice_ids=task_info.get("voice_ids", []),
                 custom_title=task_info.get("custom_title", ""),
+                title_config=task_info.get("title_config", {}),
             )
 
             if gen_task:
@@ -1430,51 +1745,47 @@ def generate_video(self, task_id: str) -> dict:
 
             _update_task_progress(task_id, 95, "上传完成")
 
-            # ── 4.5 封面抽帧 ────────────────────────────────────────────────
-            # 预览视频上传完成后，提取封面帧写入 gen_task.cover_url
-            # 这样封面路由（generation_cover.py 步骤A）可以通过 generation_task_id 直接找到
+            # ── 4.5 封面帧持久化 ────────────────────────────────────────────
+            # RenderAdapter 在渲染完成后已用本地 ffmpeg 从 output_path 抽帧
+            # （标题通过 ASS 烧录，帧天然带标题），并上传 OSS 返回 cover_candidates。
+            # 这里把第一帧写入 gen_task.cover_url，完整列表写入 metadata，
+            # 封面路由（generation_cover.py）的 A/B/C/D 步骤即可直接命中。
             try:
-                from packages.shared.mediakit_client import get_mediakit_client
+                if cover_candidates:
+                    first = cover_candidates[0]
+                    # 候选帧字段兼容：RenderAdapter 用 image_url，thumbnail_generator 用 url
+                    cover_frame_url = first.get("image_url") or first.get("url") or ""
+                    if cover_frame_url:
+                        _cover_session = SessionLocal()
+                        try:
+                            from packages.adapters.sqlalchemy_impl.models import (
+                                GenerationTaskModel,
+                            )
 
-                mk_client = get_mediakit_client()
-                if mk_client.is_available:
-                    _update_task_progress(task_id, 96, "提取封面帧")
-                    snapshots = mk_client.extract_frames(
-                        video_url=file_url,
-                        strategy="SpecifiedFrames",
-                        max_frames=1,
-                    )
-                    if snapshots and len(snapshots) > 0:
-                        cover_frame_url = snapshots[0].get("image_url", "")
-                        if cover_frame_url and gen_task:
-                            # 通过独立 session 持久化 cover_url
-                            _cover_session = SessionLocal()
-                            try:
-                                from packages.adapters.sqlalchemy_impl.models import (
-                                    GenerationTaskModel,
+                            _cover_model = (
+                                _cover_session.query(GenerationTaskModel)
+                                .filter(GenerationTaskModel.id == task_id)
+                                .first()
+                            )
+                            if _cover_model:
+                                _cover_model.cover_url = cover_frame_url
+                                # 持久化完整候选列表到 metadata
+                                meta = dict(_cover_model.metadata or {})
+                                meta["cover_candidates"] = cover_candidates
+                                _cover_model.metadata = meta
+                                _cover_session.commit()
+                                logger.info(
+                                    "[task_id=%s] 封面帧已持久化(ffmpeg本地抽帧): cover_url=%s candidates=%d",
+                                    task_id,
+                                    cover_frame_url[:80],
+                                    len(cover_candidates),
                                 )
-
-                                _cover_model = (
-                                    _cover_session.query(GenerationTaskModel)
-                                    .filter(GenerationTaskModel.id == task_id)
-                                    .first()
-                                )
-                                if _cover_model:
-                                    _cover_model.cover_url = cover_frame_url
-                                    _cover_session.commit()
-                                    logger.info(
-                                        "[task_id=%s] 封面帧提取成功: %s",
-                                        task_id,
-                                        cover_frame_url[:80],
-                                    )
-                            finally:
-                                _cover_session.close()
-                    else:
-                        logger.warning("[task_id=%s] 封面帧提取返回空结果", task_id)
+                        finally:
+                            _cover_session.close()
                 else:
-                    logger.warning("[task_id=%s] MediaKit 未配置，跳过封面帧提取", task_id)
+                    logger.warning("[task_id=%s] 渲染未产出 cover_candidates，封面将依赖 API 兜底", task_id)
             except Exception:
-                logger.warning("[task_id=%s] 封面帧提取失败（不影响主流程）", task_id, exc_info=True)
+                logger.warning("[task_id=%s] 封面帧持久化失败（不影响主流程）", task_id, exc_info=True)
 
             # ── 5. 标记完成 ──────────────────────────────────────────────────
             _update_task_status(task_id, "mark_completed", result_count=video_count)

infra/docker/api-base.Dockerfile

@@ -17,6 +17,8 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     libpq-dev \
     libpq5 \
     ffmpeg \
+    fonts-noto-cjk \
+    fontconfig \
     && rm -rf /var/lib/apt/lists/*
 
 # 创建虚拟环境

infra/docker/entrypoint-worker.sh

@@ -6,8 +6,13 @@ set -e
 
 CONCURRENCY="${WORKER_CONCURRENCY:-2}"
 
+# ⚠️ 部署约束：此 Worker 必须且只能运行单实例（replicas=1）
+# -B 标志嵌入 celery beat，beat 负责定期触发 pending 超时清理等定时任务
+# 多实例部署会导致每个 Worker 独立运行 Beat，造成定时任务重复执行
+# 若需横向扩展 Worker，必须将 Beat 拆分为独立服务（celery beat -A worker_app.celery_app）
 exec celery \
   -A worker_app.celery_app \
   worker \
   --loglevel=info \
+  "-B" \
   "--concurrency=${CONCURRENCY}"

packages/adapters/sqlalchemy_impl/generation_task_repository.py

@@ -42,6 +42,7 @@ def _to_domain(model: GenerationTaskModel) -> GenerationTask:
         output_height=getattr(model, "output_height", 720) or 720,
         cover_url=getattr(model, "cover_url", "") or "",
         custom_title=getattr(model, "custom_title", "") or "",
+        title_config=dict(getattr(model, "title_config", {}) or {}),
         logs=model.logs or "[]",
         created_at=model.created_at,
         updated_at=model.updated_at,
@@ -86,6 +87,7 @@ class SQLAlchemyGenerationTaskRepository:
             output_height=task.output_height,
             cover_url=task.cover_url or "",
             custom_title=task.custom_title or "",
+            title_config=dict(task.title_config) if task.title_config else {},
             logs=task.logs,
             created_at=task.created_at,
             updated_at=task.updated_at,
@@ -273,6 +275,7 @@ class SQLAlchemyGenerationTaskRepository:
         model.output_height = task.output_height
         model.cover_url = task.cover_url or ""
         model.custom_title = task.custom_title or ""
+        model.title_config = dict(task.title_config) if task.title_config else {}
         model.logs = task.logs
         self.session.commit()
         return task
@@ -310,3 +313,42 @@ class SQLAlchemyGenerationTaskRepository:
             model.completed_at = datetime.now(timezone.utc)
         self.session.commit()
         return len(models)
+
+    def cleanup_stale_pending(self, timeout_minutes: int = 30) -> int:
+        """清理超时的 pending 任务（未被 Worker 拉取的任务）。
+
+        全局任务队列有 pending 数量上限，长期卡在 pending 的任务会占满队列，
+        导致新用户无法创建任务。将超时的 pending 任务标记为 failed。
+
+        Args:
+            timeout_minutes: 超时时间（分钟），默认 30 分钟
+
+        Returns:
+            清理的任务数量
+        """
+        from datetime import timedelta
+
+        cutoff = datetime.now(timezone.utc) - timedelta(minutes=timeout_minutes)
+        error_info = {
+            "error_type": "PendingTimeout",
+            "message": f"任务在 pending 状态停留超过 {timeout_minutes} 分钟，自动清理",
+            "failed_at": datetime.now(timezone.utc).isoformat(),
+        }
+        count = (
+            self.session.query(GenerationTaskModel)
+            .filter(
+                GenerationTaskModel.status == GenerationTaskStatus.PENDING.value,
+                GenerationTaskModel.created_at < cutoff,
+            )
+            .update(
+                {
+                    GenerationTaskModel.status: GenerationTaskStatus.FAILED.value,
+                    GenerationTaskModel.error_message: "pending timeout: auto cleanup",
+                    GenerationTaskModel.error_info: error_info,
+                    GenerationTaskModel.completed_at: datetime.now(timezone.utc),
+                },
+                synchronize_session=False,
+            )
+        )
+        self.session.commit()
+        return count

packages/adapters/sqlalchemy_impl/models.py

@@ -298,6 +298,7 @@ class GenerationTaskModel(Base):
     output_height = Column(Integer, nullable=False, default=720)
     cover_url = Column(String(1000), nullable=False, default="")
     custom_title = Column(String(500), nullable=False, default="")
+    title_config = Column(JSON, nullable=False, default=dict)
     bgm_config = Column(JSON, nullable=False, default=dict)
     extra_meta = Column("metadata", JSON, nullable=False, default=dict)
     logs = Column(Text, nullable=False, default="[]", server_default="[]")

packages/application/generation_tasks.py

@@ -69,6 +69,7 @@ class CreateGenerationTaskUseCase:
             output_height=command.output_height,
             cover_url=command.cover_url,
             custom_title=command.custom_title,
+            title_config=command.title_config,
         )
         return self.generation_task_repository.create(task)
 

packages/domain/generation_task.py

@@ -121,6 +121,7 @@ class GenerationTask:
     output_height: int = 720
     cover_url: str = ""
     custom_title: str = ""
+    title_config: dict = field(default_factory=dict)
     extra_meta: dict = field(default_factory=dict)
     logs: str = "[]"
     created_at: datetime = field(default_factory=lambda: datetime.now(timezone.utc))
@@ -153,6 +154,7 @@ class GenerationTask:
         output_height: int = 720,
         cover_url: str = "",
         custom_title: str = "",
+        title_config: dict | None = None,
         extra_meta: dict | None = None,
     ) -> "GenerationTask":
         if not project_id.strip() and not template_id.strip():
@@ -184,6 +186,7 @@ class GenerationTask:
             output_height=output_height,
             cover_url=cover_url,
             custom_title=custom_title,
+            title_config=dict(title_config) if title_config else {},
             extra_meta=dict(extra_meta) if extra_meta else {},
         )
 

packages/shared/mediakit_client.py

@@ -364,7 +364,7 @@ class MediaKitClient:
                 data = response.json()
 
                 status = data.get("status")
-                if status == "success":
+                if status in ("completed", "success"):
                     result = data.get("result", {})
                     snapshots = result.get("snapshots", [])
                     logger.info(

packages/shared/title_overlay.py

@@ -0,0 +1,171 @@
+"""封面标题文字叠加（Pillow）— API / Worker 共用。
+
+在封面帧上绘制白色标题文字 + 黑色描边/阴影，支持 CJK 字体和自动换行。
+从已渲染视频抽帧时通常不需要调用（标题已烧录）；
+从源素材抽帧（API E2 兜底）时调用，保证封面带标题。
+"""
+
+from __future__ import annotations
+
+import logging
+from pathlib import Path
+from typing import Optional
+
+logger = logging.getLogger(__name__)
+
+# 按优先级查找 CJK 字体（Debian/Ubuntu fonts-noto-cjk 安装路径）
+_FONT_CANDIDATES = (
+    "/usr/share/fonts/opentype/noto/NotoSansCJK-Bold.ttc",
+    "/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc",
+    "/usr/share/fonts/truetype/noto/NotoSansCJK-Bold.ttc",
+    "/usr/share/fonts/truetype/noto/NotoSansCJK-Regular.ttc",
+    "/usr/share/fonts/truetype/wqy/wqy-zenhei.ttc",
+)
+
+
+def find_title_font(size: int):
+    """查找可用的 CJK 字体并返回 PIL ImageFont，找不到返回 None。"""
+    try:
+        from PIL import ImageFont
+    except ImportError:
+        return None
+
+    for fp in _FONT_CANDIDATES:
+        if Path(fp).exists():
+            try:
+                return ImageFont.truetype(fp, size=size)
+            except Exception:
+                continue
+    logger.warning("未找到 CJK 字体，标题叠加将使用 PIL 默认字体（中文可能显示为方块）")
+    return ImageFont.load_default()
+
+
+def _parse_hex_color(color: str, fallback=(255, 255, 255)) -> tuple[int, int, int]:
+    "将 #RRGGBB / #RGB 解析为 RGB 元组，失败返回 fallback。"
+    if not color or not isinstance(color, str):
+        return fallback
+    c = color.strip().lstrip("#")
+    try:
+        if len(c) == 6:
+            return (int(c[0:2], 16), int(c[2:4], 16), int(c[4:6], 16))
+        if len(c) == 3:
+            return (int(c[0] * 2, 16), int(c[1] * 2, 16), int(c[2] * 2, 16))
+    except (ValueError, IndexError):
+        pass
+    return fallback
+
+
+def wrap_title_text(text: str, font, max_width: int) -> list[str]:
+    """按像素宽度对中英文混合文本自动换行，支持显式 \\n。"""
+    lines: list[str] = []
+    current = ""
+    for ch in text:
+        if ch == "\n":
+            if current:
+                lines.append(current)
+            current = ""
+            continue
+        trial = current + ch
+        try:
+            bbox = font.getbbox(trial)
+            width = bbox[2] - bbox[0]
+        except Exception:
+            width = len(trial) * (font.size // 2)
+        if width <= max_width:
+            current = trial
+        else:
+            if current:
+                lines.append(current)
+            current = ch
+    if current:
+        lines.append(current)
+    return lines
+
+
+def apply_title_to_image(
+    image_path: str,
+    title_text: str,
+    *,
+    color: str = "#ffffff",
+    position: str = "bottom",
+    font_size: Optional[int] = None,
+    margin_ratio: float = 0.06,
+    stroke_width_ratio: float = 0.04,
+) -> Optional[str]:
+    """在图片上绘制标题文字并覆盖保存。
+
+    Args:
+        image_path: 图片路径（处理结果覆盖写回）
+        title_text: 标题文字；为空直接返回 None 表示跳过
+        color: 字体颜色（#RRGGBB），默认白色
+        position: top / center / bottom
+        font_size: 字号，None 时按图片宽度自动计算
+        margin_ratio: 边缘留白占短边比例
+        stroke_width_ratio: 描边宽度占字号比例
+
+    Returns:
+        成功返回 image_path；标题为空或 PIL 不可用返回 None。
+    """
+    if not title_text or not title_text.strip():
+        return None
+
+    try:
+        from PIL import Image, ImageDraw
+    except ImportError:
+        logger.warning("Pillow 未安装，跳过标题叠加: image=%s", image_path)
+        return None
+
+    img = Image.open(image_path).convert("RGB")
+    draw = ImageDraw.Draw(img)
+    img_w, img_h = img.size
+
+    if font_size is None:
+        font_size = max(28, min(72, img_w // 16))
+
+    font = find_title_font(font_size)
+    if font is None:
+        return None
+
+    text_rgb = _parse_hex_color(color)
+    stroke_width = max(2, int(font_size * stroke_width_ratio))
+    margin = int(min(img_w, img_h) * margin_ratio)
+    max_text_width = img_w - 2 * margin
+
+    lines = wrap_title_text(title_text.strip(), font, max_text_width)
+    if not lines:
+        return None
+
+    line_heights = []
+    for ln in lines:
+        bbox = font.getbbox(ln)
+        line_heights.append(bbox[3] - bbox[1])
+    line_height = max(line_heights) if line_heights else font_size
+    line_gap = int(line_height * 0.3)
+    total_height = len(lines) * line_height + (len(lines) - 1) * line_gap
+
+    if position == "top":
+        y_start = margin
+    elif position == "center":
+        y_start = (img_h - total_height) // 2
+    else:
+        y_start = img_h - total_height - margin
+
+    for i, ln in enumerate(lines):
+        bbox = font.getbbox(ln)
+        line_w = bbox[2] - bbox[0]
+        x = (img_w - line_w) // 2
+        y = y_start + i * (line_height + line_gap)
+        # 阴影
+        draw.text((x + 2, y + 2), ln, font=font, fill=(0, 0, 0))
+        # 文字（颜色由 color 参数控制）+ 黑色描边
+        draw.text(
+            (x, y),
+            ln,
+            font=font,
+            fill=text_rgb,
+            stroke_width=stroke_width,
+            stroke_fill=(0, 0, 0),
+        )
+
+    img.save(image_path, "JPEG", quality=92)
+    return image_path

requirements-base.txt

@@ -29,3 +29,4 @@ httpx==0.27.2
 
 # Prometheus monitoring
 prometheus-client==0.21.1
+Pillow==10.4.0

tests/unit/test_1294_preview_voice_injection.py

@@ -158,7 +158,7 @@ class TestRenderVideoVoiceInjection:
 
             from packages.domain import EditingMode
 
-            output_path, render_duration = _render_video(
+            output_path, render_duration, cover_candidates = _render_video(
                 task_id="test_task_123",
                 downloaded_videos=[Path("/tmp/video1.mp4")],
                 voice_path=None,

tests/unit/test_batch_clips_endpoint.py

@@ -0,0 +1,135 @@
+"""Tests for PUT /templates/{id}/editor/clips batch update endpoint.
+
+Updated for transactional replace_all_clips_transactional method.
+"""
+
+from __future__ import annotations
+
+import os
+from unittest.mock import MagicMock
+
+import pytest
+
+os.environ.setdefault("JWT_SECRET_KEY", "unit-test-secret-key-for-testing")
+
+
+@pytest.fixture
+def mock_services():
+    plan_svc = MagicMock()
+    tpl_svc = MagicMock()
+    plan_svc.get_plan_or_raise.return_value = MagicMock(id="plan-1", template_id="tpl-1")
+    plan_svc.replace_all_clips_transactional.return_value = 2
+    return tpl_svc, plan_svc
+
+
+class TestBatchUpdateClips:
+    def test_batch_update_calls_transactional_replace(self, mock_services):
+        """验证批量更新调用事务性替换方法，传入正确的参数。"""
+        from app.api.routes.templates_editor.draft import batch_update_clips
+        from app.api.routes.templates_editor.schemas import (
+            EditorClipBatchItem,
+            EditorClipBatchUpdateRequest,
+        )
+
+        _, plan_svc = mock_services
+        req = EditorClipBatchUpdateRequest(
+            clips=[
+                EditorClipBatchItem(asset_id="a1", start_time=0.0, duration=3.0, order=0),
+                EditorClipBatchItem(asset_id="a2", start_time=3.0, duration=5.0, order=1),
+            ]
+        )
+
+        result = batch_update_clips(
+            template_id="tpl-1",
+            req=req,
+            plan_id="plan-1",
+            services=mock_services,
+            _=MagicMock(),
+        )
+
+        assert result.plan_id == "plan-1"
+        assert result.clip_count == 2
+        plan_svc.replace_all_clips_transactional.assert_called_once()
+        call_args = plan_svc.replace_all_clips_transactional.call_args
+        assert call_args[0][0] == "plan-1"
+        clips_data = call_args[0][1]
+        assert len(clips_data) == 2
+        assert clips_data[0]["asset_id"] == "a1"
+        assert clips_data[0]["start_time"] == 0.0
+        assert clips_data[0]["duration"] == 3.0
+        assert clips_data[1]["asset_id"] == "a2"
+
+    def test_batch_update_empty_clips(self, mock_services):
+        """空 clips 列表也能正常处理。"""
+        from app.api.routes.templates_editor.draft import batch_update_clips
+        from app.api.routes.templates_editor.schemas import EditorClipBatchUpdateRequest
+
+        _, plan_svc = mock_services
+        req = EditorClipBatchUpdateRequest(clips=[])
+
+        result = batch_update_clips(
+            template_id="tpl-1",
+            req=req,
+            plan_id="plan-1",
+            services=mock_services,
+            _=MagicMock(),
+        )
+
+        assert result.clip_count == 0
+        plan_svc.replace_all_clips_transactional.assert_called_once()
+        call_args = plan_svc.replace_all_clips_transactional.call_args
+        assert call_args[0][1] == []
+
+    def test_batch_update_passes_order_correctly(self, mock_services):
+        """验证 order 字段正确传递。"""
+        from app.api.routes.templates_editor.draft import batch_update_clips
+        from app.api.routes.templates_editor.schemas import (
+            EditorClipBatchItem,
+            EditorClipBatchUpdateRequest,
+        )
+
+        _, plan_svc = mock_services
+        req = EditorClipBatchUpdateRequest(
+            clips=[
+                EditorClipBatchItem(asset_id="a1", start_time=0.0, duration=3.0, order=5),
+            ]
+        )
+
+        batch_update_clips(
+            template_id="tpl-1",
+            req=req,
+            plan_id="plan-1",
+            services=mock_services,
+            _=MagicMock(),
+        )
+
+        clips_data = plan_svc.replace_all_clips_transactional.call_args[0][1]
+        assert clips_data[0]["order"] == 5
+        assert clips_data[0]["asset_id"] == "a1"
+        assert clips_data[0]["start_time"] == 0.0
+        assert clips_data[0]["duration"] == 3.0
+
+
+class TestEditorClipBatchItemValidation:
+    """验证 schema 校验规则。"""
+
+    def test_asset_id_empty_string_allowed(self):
+        """asset_id 空字符串允许通过（占位片段场景）。"""
+        from app.api.routes.templates_editor.schemas import EditorClipBatchItem
+
+        item = EditorClipBatchItem(asset_id="", start_time=0.0, duration=3.0, order=0)
+        assert item.asset_id == ""
+
+    def test_asset_id_valid(self):
+        """有效 asset_id 应通过校验。"""
+        from app.api.routes.templates_editor.schemas import EditorClipBatchItem
+
+        item = EditorClipBatchItem(asset_id="abc123", start_time=0.0, duration=3.0, order=0)
+        assert item.asset_id == "abc123"
+
+    def test_order_none_by_default(self):
+        """order 默认为 None，表示按数组顺序。"""
+        from app.api.routes.templates_editor.schemas import EditorClipBatchItem
+
+        item = EditorClipBatchItem(asset_id="a1", start_time=0.0, duration=3.0)
+        assert item.order is None

tests/unit/test_cover_url_finalize.py

@@ -1,132 +0,0 @@
-"""Tests for cover_url backfill to GenerationTask.
-
-Verifies _finalize_render_success correctly writes cover_url
-from cover_candidates to gen_task.cover_url.
-"""
-
-from __future__ import annotations
-
-import sys
-from pathlib import Path
-from unittest.mock import MagicMock, patch
-
-import pytest
-
-# Add worker app to sys.path
-_WORKER_ROOT = Path(__file__).resolve().parents[2] / "apps" / "worker"
-if str(_WORKER_ROOT) not in sys.path:
-    sys.path.insert(0, str(_WORKER_ROOT))
-
-
-class FakeGenTask:
-    """Simple stand-in for GenerationTask that tracks attribute assignment."""
-
-    def __init__(self):
-        object.__setattr__(self, "_assigned", {})
-        self.id = "task-1"
-        self.status = MagicMock()
-        self.status.value = "running"
-
-    def __setattr__(self, name, value):
-        if not name.startswith("_"):
-            self._assigned[name] = value
-        object.__setattr__(self, name, value)
-
-    def append_log(self, **kwargs):
-        pass
-
-
-def _make_plan():
-    plan = MagicMock()
-    plan.project_id = "proj-1"
-    plan.created_by_user_id = "user-1"
-    plan.config = {"batch_id": "batch-1", "mode": "edit_plan", "title": {"text": "test"}}
-    plan.mark_completed = MagicMock()
-    return plan
-
-
-def _call_finalize(cover_candidates=None, gen_task=None, plan=None):
-    from worker_app.tasks.edit_plan_generation import _finalize_render_success
-
-    plan = plan or _make_plan()
-    gen_task = gen_task or FakeGenTask()
-
-    plan_repo = MagicMock()
-    clip_repo = MagicMock()
-    gen_task_repo = MagicMock()
-    gen_task_repo.get.return_value = gen_task
-    db = MagicMock()
-
-    with patch("worker_app.tasks.edit_plan_generation.create_video_record_and_dedup"):
-        result = _finalize_render_success(
-            plan=plan,
-            plan_repo=plan_repo,
-            clip_repo=clip_repo,
-            gen_task_repo=gen_task_repo,
-            db=db,
-            plan_id="plan-1",
-            output_url="https://oss.example.com/output.mp4",
-            storage_key="rendered/plan-1/task-1.mp4",
-            duration=10.0,
-            file_size=1024,
-            width=1280,
-            height=720,
-            rendered_clip_ids=["clip-1"],
-            failed_clip_ids=[],
-            generation_task_id="task-1",
-            output_path=Path("/tmp/output.mp4"),
-            engine="unified",
-            thumbnail_url="",
-            cover_candidates=cover_candidates,
-        )
-
-    return result, gen_task, gen_task_repo
-
-
-class TestFinalizeCoverUrl:
-
-    def test_cover_url_set_from_image_url(self):
-        """cover_candidates with image_url should set gen_task.cover_url"""
-        candidates = [
-            {"image_url": "https://oss.example.com/cover1.jpg", "frame_time": 1.5},
-            {"image_url": "https://oss.example.com/cover2.jpg", "frame_time": 3.0},
-        ]
-        _, gen_task, gen_task_repo = _call_finalize(cover_candidates=candidates)
-        assert gen_task.cover_url == "https://oss.example.com/cover1.jpg"
-        gen_task_repo.update.assert_called()
-
-    def test_cover_url_fallback_to_url_key(self):
-        """Should fallback to 'url' key when 'image_url' is absent"""
-        candidates = [{"url": "https://oss.example.com/cover_url_key.jpg"}]
-        _, gen_task, _ = _call_finalize(cover_candidates=candidates)
-        assert gen_task.cover_url == "https://oss.example.com/cover_url_key.jpg"
-
-    def test_cover_url_not_set_when_empty_list(self):
-        """Empty cover_candidates should not set cover_url"""
-        _, gen_task, _ = _call_finalize(cover_candidates=[])
-        assert "cover_url" not in gen_task._assigned
-
-    def test_cover_url_not_set_when_none(self):
-        """None cover_candidates should not set cover_url"""
-        _, gen_task, _ = _call_finalize(cover_candidates=None)
-        assert "cover_url" not in gen_task._assigned
-
-    def test_cover_url_not_set_when_url_empty(self):
-        """Empty URL strings in candidates should not set cover_url"""
-        candidates = [{"image_url": "", "url": ""}]
-        _, gen_task, _ = _call_finalize(cover_candidates=candidates)
-        assert "cover_url" not in gen_task._assigned
-
-    def test_no_generation_task_no_crash(self):
-        """Should not crash when gen_task is None"""
-        candidates = [{"image_url": "https://oss.example.com/cover.jpg"}]
-        gen_task_repo = MagicMock()
-        gen_task_repo.get.return_value = None
-        result, _, _ = _call_finalize(cover_candidates=candidates)
-        assert result["status"] == "completed"
-
-    def test_image_url_priority_over_url(self):
-        """image_url should take priority over url key"""
-        candidates = [{"image_url": "https://a.jpg", "url": "https://b.jpg"}]
-        _, gen_task, _ = _call_finalize(cover_candidates=candidates)
-        assert gen_task.cover_url == "https://a.jpg"

tests/unit/test_edit_plan_worker_failure.py

@@ -1,333 +0,0 @@
-"""P0-2: Celery 任务 render_edit_plan 失败时更新 GenerationTask 状态。
-
-验证：
-  - 异常发生时 GenerationTask 状态更新为 failed
-  - error_message 记录了异常类型和描述
-  - completed_at 被设置
-  - 即使 generation_task_id 为空也不崩溃
-  - 即使更新 GenerationTask 本身失败也不影响 retry
-"""
-
-from __future__ import annotations
-
-import os
-import sys
-from dataclasses import dataclass, field
-from types import ModuleType
-from typing import Any, Optional
-from unittest.mock import MagicMock, patch
-
-os.environ.setdefault("JWT_SECRET_KEY", "unit-test-secret-key-for-testing")
-os.environ.setdefault("DATABASE_URL", "sqlite:///test.db")
-
-import pytest
-
-# ── Mock worker 模块以避免数据库连接 ──────────────────────────────────────────
-# worker_app.db 在 import 时会尝试连接数据库，必须在导入 task 模块前 mock
-
-sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", "..", "apps", "worker"))
-sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", ".."))
-
-# 预注册 mock 模块，阻止真实数据库初始化
-_mock_db_mod = ModuleType("worker_app.db")
-_mock_db_mod.SessionLocal = MagicMock()
-sys.modules.setdefault("worker_app.db", _mock_db_mod)
-
-_mock_celery_mod = ModuleType("worker_app.celery_app")
-_mock_celery_app = MagicMock()
-# 让 @celery_app.task(...) 装饰器透传原始函数，否则函数变成 MagicMock
-_mock_celery_app.task = lambda **kwargs: lambda fn: fn
-_mock_celery_mod.celery_app = _mock_celery_app
-sys.modules.setdefault("worker_app.celery_app", _mock_celery_mod)
-
-
-# ── Stub domain objects ───────────────────────────────────────────────────────
-
-
-@dataclass
-class _StubStatus:
-    value: str
-
-    def __eq__(self, other):
-        if isinstance(other, str):
-            return self.value == other
-        if isinstance(other, _StubStatus):
-            return self.value == other.value
-        return NotImplemented
-
-
-@dataclass
-class StubEditPlan:
-    id: str = "plan-001"
-    template_id: str = "tmpl-001"
-    status: Any = None
-    config: dict = field(default_factory=dict)
-    project_id: str = ""
-    created_by_user_id: str = "user-001"
-
-    def mark_failed(self):
-        self.status = _StubStatus("failed")
-
-    def mark_completed(self):
-        self.status = _StubStatus("completed")
-
-
-@dataclass
-class StubGenerationTask:
-    id: str = "gen-task-001"
-    status: Any = field(default_factory=lambda: _StubStatus("pending"))
-    error_message: str = ""
-    progress: float = 0.0
-    result_count: int = 0
-    started_at: Any = None
-    completed_at: Any = None
-    project_id: str = ""
-    created_by_user_id: str = "user-001"
-
-
-@dataclass
-class StubClip:
-    id: str = "clip-001"
-    plan_id: str = "plan-001"
-    asset_id: str = "assets/video.mp4"
-    order: int = 1
-    status: Any = field(default_factory=lambda: _StubStatus("ready"))
-    transition_effect: str = ""
-    text_content: str = ""
-    clip_type: str = "MAIN"
-    duration: float = 0.0
-
-    def mark_failed(self):
-        self.status = _StubStatus("failed")
-
-    def mark_rendered(self):
-        self.status = _StubStatus("rendered")
-
-
-# ── Stub repositories ─────────────────────────────────────────────────────────
-
-
-class StubPlanRepo:
-    def __init__(self, plan: StubEditPlan):
-        self._plan = plan
-
-    def get(self, plan_id: str) -> Optional[StubEditPlan]:
-        if plan_id == self._plan.id:
-            return self._plan
-        return None
-
-    def update(self, plan: StubEditPlan) -> StubEditPlan:
-        self._plan = plan
-        return plan
-
-
-class StubClipRepo:
-    def __init__(self, clips: list[StubClip] | None = None):
-        self._clips = clips or []
-
-    def list_by_plan(self, plan_id: str, skip: int = 0, limit: int = 10000) -> list[StubClip]:
-        return [c for c in self._clips if c.plan_id == plan_id]
-
-    def get(self, clip_id: str) -> Optional[StubClip]:
-        for c in self._clips:
-            if c.id == clip_id:
-                return c
-        return None
-
-    def update(self, clip: StubClip) -> StubClip:
-        return clip
-
-
-class StubGenTaskRepo:
-    def __init__(self, task: StubGenerationTask | None = None):
-        self._store: dict[str, StubGenerationTask] = {}
-        if task:
-            self._store[task.id] = task
-
-    def get(self, task_id: str) -> Optional[StubGenerationTask]:
-        return self._store.get(task_id)
-
-    def update(self, task: StubGenerationTask) -> StubGenerationTask:
-        self._store[task.id] = task
-        return task
-
-
-# ── Import task module (after mocks are in place) ─────────────────────────────
-
-from worker_app.tasks.edit_plan_generation import render_edit_plan
-
-# ── Tests ─────────────────────────────────────────────────────────────────────
-
-
-class TestRenderEditPlanFailureUpdatesGenTask:
-    """P0-2: render_edit_plan 异常时更新 GenerationTask 状态为 failed"""
-
-    def _make_bound_task(self):
-        """构建绑定的 Celery task mock"""
-        task = MagicMock()
-        task.retry = MagicMock(side_effect=RuntimeError("retry called"))
-        return task
-
-    def test_exception_marks_gen_task_failed(self):
-        """异常时 GenerationTask.status 被设为 failed"""
-        plan = StubEditPlan(status=_StubStatus("rendering"))
-        plan.config["generation_task_id"] = "gen-task-001"
-        gen_task = StubGenerationTask(id="gen-task-001", status=_StubStatus("running"))
-
-        plan_repo = StubPlanRepo(plan)
-        gen_task_repo = StubGenTaskRepo(gen_task)
-
-        # 让 clip_repo 抛异常以触发 except 路径
-        clip_repo_bad = MagicMock()
-        clip_repo_bad.list_by_plan.side_effect = RuntimeError("OSS 连接失败")
-
-        def fake_get_repos():
-            yield plan_repo, clip_repo_bad, gen_task_repo, MagicMock()
-
-        bound_task = self._make_bound_task()
-
-        with patch(
-            "worker_app.tasks.edit_plan_generation._get_repos",
-            side_effect=fake_get_repos,
-        ):
-            with pytest.raises(RuntimeError, match="retry called"):
-                render_edit_plan(bound_task, "plan-001")
-
-        # 核心断言：GenerationTask 状态为 failed（生产代码赋值为字符串）
-        assert gen_task.status == "failed"
-
-    def test_exception_records_error_message(self):
-        """异常时 error_message 包含异常类型和描述"""
-        plan = StubEditPlan(status=_StubStatus("rendering"))
-        plan.config["generation_task_id"] = "gen-task-001"
-        gen_task = StubGenerationTask(id="gen-task-001", status=_StubStatus("running"))
-
-        plan_repo = StubPlanRepo(plan)
-        clip_repo = MagicMock()
-        clip_repo.list_by_plan.side_effect = RuntimeError("DB 查询超时")
-        gen_task_repo = StubGenTaskRepo(gen_task)
-
-        def fake_get_repos():
-            yield plan_repo, clip_repo, gen_task_repo, MagicMock()
-
-        bound_task = self._make_bound_task()
-
-        with patch(
-            "worker_app.tasks.edit_plan_generation._get_repos",
-            side_effect=fake_get_repos,
-        ):
-            with pytest.raises(RuntimeError, match="retry called"):
-                render_edit_plan(bound_task, "plan-001")
-
-        assert gen_task.status == "failed"
-        assert "DB 查询超时" in gen_task.error_message
-        assert "RuntimeError" in gen_task.error_message
-
-    def test_exception_sets_completed_at(self):
-        """异常时 completed_at 被设置"""
-        plan = StubEditPlan(status=_StubStatus("rendering"))
-        plan.config["generation_task_id"] = "gen-task-001"
-        gen_task = StubGenerationTask(id="gen-task-001", status=_StubStatus("running"))
-
-        plan_repo = StubPlanRepo(plan)
-        clip_repo = MagicMock()
-        clip_repo.list_by_plan.side_effect = RuntimeError("boom")
-        gen_task_repo = StubGenTaskRepo(gen_task)
-
-        def fake_get_repos():
-            yield plan_repo, clip_repo, gen_task_repo, MagicMock()
-
-        bound_task = self._make_bound_task()
-
-        with patch(
-            "worker_app.tasks.edit_plan_generation._get_repos",
-            side_effect=fake_get_repos,
-        ):
-            with pytest.raises(RuntimeError):
-                render_edit_plan(bound_task, "plan-001")
-
-        assert gen_task.completed_at is not None
-
-    def test_no_generation_task_id_does_not_crash(self):
-        """generation_task_id 为空时，异常处理不崩溃"""
-        plan = StubEditPlan(status=_StubStatus("rendering"))
-        plan.config = {}  # 不设置 generation_task_id
-
-        plan_repo = StubPlanRepo(plan)
-        clip_repo = MagicMock()
-        clip_repo.list_by_plan.side_effect = RuntimeError("boom")
-        gen_task_repo = StubGenTaskRepo()  # 空 repo
-
-        def fake_get_repos():
-            yield plan_repo, clip_repo, gen_task_repo, MagicMock()
-
-        bound_task = self._make_bound_task()
-
-        with patch(
-            "worker_app.tasks.edit_plan_generation._get_repos",
-            side_effect=fake_get_repos,
-        ):
-            with pytest.raises(RuntimeError, match="retry called"):
-                render_edit_plan(bound_task, "plan-001")
-
-        # 计划仍被标记为 failed
-        assert plan.status.value == "failed"
-
-    def test_gen_task_update_failure_does_not_block_retry(self):
-        """更新 GenerationTask 失败时，不影响 retry 流程"""
-        plan = StubEditPlan(status=_StubStatus("rendering"))
-        plan.config["generation_task_id"] = "gen-task-001"
-        gen_task = StubGenerationTask(id="gen-task-001", status=_StubStatus("running"))
-
-        plan_repo = StubPlanRepo(plan)
-        clip_repo = MagicMock()
-        clip_repo.list_by_plan.side_effect = RuntimeError("原始错误")
-        # gen_task_repo.update 也抛异常
-        gen_task_repo = MagicMock()
-        gen_task_repo.get.return_value = gen_task
-        gen_task_repo.update.side_effect = RuntimeError("DB 写入失败")
-
-        def fake_get_repos():
-            yield plan_repo, clip_repo, gen_task_repo, MagicMock()
-
-        bound_task = self._make_bound_task()
-
-        with patch(
-            "worker_app.tasks.edit_plan_generation._get_repos",
-            side_effect=fake_get_repos,
-        ):
-            with pytest.raises(RuntimeError, match="retry called"):
-                render_edit_plan(bound_task, "plan-001")
-
-        # retry 被调用说明流程正确
-        bound_task.retry.assert_called_once()
-
-    def test_already_failed_gen_task_not_overwritten(self):
-        """已经 failed 的 GenerationTask 不会被重复更新"""
-        plan = StubEditPlan(status=_StubStatus("rendering"))
-        plan.config["generation_task_id"] = "gen-task-001"
-        gen_task = StubGenerationTask(
-            id="gen-task-001",
-            status=_StubStatus("failed"),  # 已经是 failed
-            error_message="之前的错误",
-        )
-
-        plan_repo = StubPlanRepo(plan)
-        clip_repo = MagicMock()
-        clip_repo.list_by_plan.side_effect = RuntimeError("新错误")
-        gen_task_repo = StubGenTaskRepo(gen_task)
-
-        def fake_get_repos():
-            yield plan_repo, clip_repo, gen_task_repo, MagicMock()
-
-        bound_task = self._make_bound_task()
-
-        with patch(
-            "worker_app.tasks.edit_plan_generation._get_repos",
-            side_effect=fake_get_repos,
-        ):
-            with pytest.raises(RuntimeError, match="retry called"):
-                render_edit_plan(bound_task, "plan-001")
-
-        # error_message 应保持原值，不被覆盖
-        assert gen_task.error_message == "之前的错误"

tests/unit/test_generation_cover.py

@@ -182,6 +182,7 @@ class TestUnifiedCoverPipelineEndpoint:
         with (
             patch("app.api.routes.generation_cover.SQLAlchemyGenerationTaskRepository") as mock_repo_cls,
             patch("packages.shared.storage.get_shared_storage_service") as mock_storage_getter,
+            patch("packages.shared.mediakit_client.get_mediakit_client") as mock_mk_getter,
         ):
             mock_repo = MagicMock()
             mock_repo.get.return_value = mock_task
@@ -194,6 +195,13 @@ class TestUnifiedCoverPipelineEndpoint:
             mock_storage_svc.get_url.return_value = "https://oss.example.com/rendered/plan-2/video.mp4"
             mock_storage_getter.return_value = mock_storage_svc
 
+            # MediaKit 抽帧也返回 None，模拟最终失败
+            mock_mk = MagicMock()
+            mock_mk.is_available = True
+            mock_mk.extract_frames.return_value = None
+            mock_mk_getter.return_value = mock_mk
+
+            # body 不传 asset_ids，步骤 E2 不会进入
             from app.api.routes.generation_cover import generate_cover
 
             with pytest.raises(HTTPException) as exc_info:
@@ -207,7 +215,6 @@ class TestUnifiedCoverPipelineEndpoint:
                 )
 
             assert exc_info.value.status_code == 400
-            assert "封面尚未生成" in exc_info.value.detail
 
     def test_cover_url_found_via_source_edit_plan(self):
         """步骤B：通过 source_edit_plan_id 找到预览任务的 cover_url。"""
@@ -318,6 +325,157 @@ class TestUnifiedCoverPipelineEndpoint:
                 template_id="template-y",
             )
 
+    def test_cover_url_found_via_cover_candidates_image_url(self):
+        """步骤D：plan.config.cover_candidates 有 image_url 时，直接使用第一个候选封面。"""
+        from unittest.mock import MagicMock, patch
+
+        from app.api.routes.generation_cover import GenerateCoverRequest
+
+        mock_plan = MagicMock()
+        # 步骤A/B/C 都找不到，进入步骤D
+        mock_plan.config = {
+            "rendered_storage_key": "rendered/plan-z/video.mp4",  # 必须有预览视频才能通过前置检查
+            "cover_candidates": [
+                {"image_url": "https://oss.example.com/candidates/cover-1.jpg", "score": 0.95},
+                {"image_url": "https://oss.example.com/candidates/cover-2.jpg", "score": 0.80},
+            ],
+        }
+
+        mock_plan_svc = MagicMock()
+        mock_plan_svc.get_plan_or_raise.return_value = mock_plan
+        mock_template_svc = MagicMock()
+        mock_db = MagicMock()
+
+        body = GenerateCoverRequest(cover_type="ai_frame")
+
+        with (
+            patch("app.api.routes.generation_cover.SQLAlchemyGenerationTaskRepository") as mock_repo_cls,
+            patch("app.api.routes.generation_cover.normalize_plan_config") as mock_normalize,
+        ):
+            mock_repo = MagicMock()
+            mock_repo.get.return_value = None
+            mock_repo.list_by_source_edit_plan.return_value = []
+            mock_repo.list_latest_completed_preview.return_value = []
+            mock_repo_cls.return_value = mock_repo
+
+            mock_normalize.return_value = {
+                "cover": {"type": "ai_frame", "image_url": "https://oss.example.com/candidates/cover-1.jpg"}
+            }
+
+            from app.api.routes.generation_cover import generate_cover
+
+            result = generate_cover(
+                body=body,
+                template_id="template-z",
+                plan_id="plan-z",
+                services=(mock_template_svc, mock_plan_svc),
+                db=mock_db,
+                current_user=MagicMock(),
+            )
+
+            # 步骤D从 cover_candidates 第一个元素的 image_url 提取封面
+            assert result.cover["image_url"] == "https://oss.example.com/candidates/cover-1.jpg"
+            # 验证 plan.config 被更新（至少调用一次：rendered_storage_key + cover）
+            assert mock_plan_svc.update_plan_config.call_count >= 1
+
+    def test_cover_url_found_via_cover_candidates_url_key(self):
+        """步骤D：cover_candidates 用 url 键（非 image_url）时，也能正确提取。"""
+        from unittest.mock import MagicMock, patch
+
+        from app.api.routes.generation_cover import GenerateCoverRequest
+
+        mock_plan = MagicMock()
+        mock_plan.config = {
+            "rendered_storage_key": "rendered/plan-w/video.mp4",  # 必须有预览视频才能通过前置检查
+            "cover_candidates": [
+                {"url": "https://oss.example.com/candidates/alt-cover.jpg"},
+            ],
+        }
+
+        mock_plan_svc = MagicMock()
+        mock_plan_svc.get_plan_or_raise.return_value = mock_plan
+        mock_template_svc = MagicMock()
+        mock_db = MagicMock()
+
+        body = GenerateCoverRequest(cover_type="ai_frame")
+
+        with (
+            patch("app.api.routes.generation_cover.SQLAlchemyGenerationTaskRepository") as mock_repo_cls,
+            patch("app.api.routes.generation_cover.normalize_plan_config") as mock_normalize,
+        ):
+            mock_repo = MagicMock()
+            mock_repo.get.return_value = None
+            mock_repo.list_by_source_edit_plan.return_value = []
+            mock_repo.list_latest_completed_preview.return_value = []
+            mock_repo_cls.return_value = mock_repo
+
+            mock_normalize.return_value = {
+                "cover": {"type": "ai_frame", "image_url": "https://oss.example.com/candidates/alt-cover.jpg"}
+            }
+
+            from app.api.routes.generation_cover import generate_cover
+
+            result = generate_cover(
+                body=body,
+                template_id="template-w",
+                plan_id="plan-w",
+                services=(mock_template_svc, mock_plan_svc),
+                db=mock_db,
+                current_user=MagicMock(),
+            )
+
+            # 步骤D fallback 到 url 键
+            assert result.cover["image_url"] == "https://oss.example.com/candidates/alt-cover.jpg"
+
+    def test_cover_candidates_skips_non_dict_first_element(self):
+        """步骤D：cover_candidates 第一个元素不是 dict 时，安全跳过不崩溃。"""
+        from unittest.mock import MagicMock, patch
+
+        from app.api.routes.generation_cover import GenerateCoverRequest
+        from fastapi import HTTPException
+
+        mock_plan = MagicMock()
+        mock_plan.config = {
+            "rendered_storage_key": "rendered/plan-skip/video.mp4",  # 必须有预览视频才能通过前置检查
+            "cover_candidates": ["not-a-dict", 42, None],
+        }
+
+        mock_plan_svc = MagicMock()
+        mock_plan_svc.get_plan_or_raise.return_value = mock_plan
+        mock_template_svc = MagicMock()
+        mock_db = MagicMock()
+
+        body = GenerateCoverRequest(cover_type="ai_frame")
+
+        with (
+            patch("app.api.routes.generation_cover.SQLAlchemyGenerationTaskRepository") as mock_repo_cls,
+            patch("packages.shared.storage.get_shared_storage_service") as mock_storage_getter,
+        ):
+            mock_repo = MagicMock()
+            mock_repo.get.return_value = None
+            mock_repo.list_by_source_edit_plan.return_value = []
+            mock_repo.list_latest_completed_preview.return_value = []
+            mock_repo_cls.return_value = mock_repo
+
+            # storage fallback 也找不到封面
+            mock_storage_svc = MagicMock()
+            mock_storage_svc.get_url.return_value = ""
+            mock_storage_getter.return_value = mock_storage_svc
+
+            from app.api.routes.generation_cover import generate_cover
+
+            # 所有步骤都失败，应返回 400
+            with pytest.raises(HTTPException) as exc_info:
+                generate_cover(
+                    body=body,
+                    template_id="template-skip",
+                    plan_id="plan-skip",
+                    services=(mock_template_svc, mock_plan_svc),
+                    db=mock_db,
+                    current_user=MagicMock(),
+                )
+            assert exc_info.value.status_code == 400
+
 
 class TestSourceEditPlanFallback:
     """测试步骤 2.5：通过 source_edit_plan_id 查找预览视频兜底逻辑。"""
@@ -687,3 +845,358 @@ class TestUploadCoverType:
             assert result.cover["image_url"] == "https://oss.example.com/uploaded/my-cover.png"
             # 验证没有调用任何预览视频查找逻辑
             # (normalize_plan_config 是唯一被调用的外部函数)
+
+    def test_cover_extracted_from_source_asset_when_no_preview(self):
+        """步骤E2：无后端渲染产物时，直接从用户选择的视频素材抽帧。"""
+        from unittest.mock import MagicMock, patch
+
+        from app.api.routes.generation_cover import GenerateCoverRequest
+
+        mock_plan = MagicMock()
+        mock_plan.config = {}  # 无 rendered_storage_key，无 generation_task_id
+
+        mock_plan_svc = MagicMock()
+        mock_plan_svc.get_plan_or_raise.return_value = mock_plan
+        mock_template_svc = MagicMock()
+        mock_db = MagicMock()
+
+        # 模拟视频素材
+        mock_asset = MagicMock()
+        mock_asset.file_type = "video"
+        mock_asset.storage_key = "uploads/source-clip.mp4"
+
+        mock_asset_repo = MagicMock()
+        mock_asset_repo.get.return_value = mock_asset
+
+        mock_mk = MagicMock()
+        mock_mk.is_available = True
+        mock_mk.extract_frames.return_value = [{"image_url": "https://mediakit.internal/frame-abc.jpg"}]
+
+        mock_storage = MagicMock()
+        mock_storage.get_url.return_value = "https://oss.example.com/uploads/source-clip.mp4"
+
+        body = GenerateCoverRequest(
+            cover_type="ai_frame",
+            asset_ids=["asset-video-1"],
+        )
+
+        with (
+            patch("app.api.routes.generation_cover.SQLAlchemyGenerationTaskRepository") as mock_repo_cls,
+            patch(
+                "packages.adapters.sqlalchemy_impl.asset_repository.SQLAlchemyAssetRepository",
+                return_value=mock_asset_repo,
+            ),
+            patch("packages.shared.mediakit_client.get_mediakit_client", return_value=mock_mk),
+            patch("packages.shared.storage.get_shared_storage_service", return_value=mock_storage),
+            patch(
+                "app.api.routes.generation_cover._persist_cover_frame",
+                return_value="https://oss.example.com/covers/final.jpg",
+            ),
+            patch("app.api.routes.generation_cover.normalize_plan_config") as mock_normalize,
+        ):
+            mock_repo = MagicMock()
+            mock_repo.get.return_value = None
+            mock_repo.list_by_source_edit_plan.return_value = []
+            mock_repo.list_latest_completed_preview.return_value = []
+            mock_repo_cls.return_value = mock_repo
+            mock_normalize.return_value = {
+                "cover": {"type": "ai_frame", "image_url": "https://oss.example.com/covers/final.jpg"}
+            }
+
+            from app.api.routes.generation_cover import generate_cover
+
+            result = generate_cover(
+                body=body,
+                template_id="tpl-source",
+                plan_id="plan-source",
+                services=(mock_template_svc, mock_plan_svc),
+                db=mock_db,
+                current_user=MagicMock(),
+            )
+
+            assert result.cover["image_url"] == "https://oss.example.com/covers/final.jpg"
+            mock_mk.extract_frames.assert_called_once()
+            # 确保用的是源素材 URL
+            call_kwargs = mock_mk.extract_frames.call_args.kwargs
+            assert "source-clip.mp4" in call_kwargs["video_url"]
+
+    def test_e2_passes_plan_title_to_persist_for_overlay(self):
+        """步骤E2：plan.config.title.text 存在时，作为 title_text 传给 _persist_cover_frame 叠加标题。"""
+        from unittest.mock import MagicMock, patch
+
+        from app.api.routes.generation_cover import GenerateCoverRequest
+
+        mock_plan = MagicMock()
+        mock_plan.config = {"title": {"enabled": True, "text": "我的视频标题"}}
+
+        mock_plan_svc = MagicMock()
+        mock_plan_svc.get_plan_or_raise.return_value = mock_plan
+        mock_template_svc = MagicMock()
+        mock_db = MagicMock()
+
+        mock_asset = MagicMock()
+        mock_asset.file_type = "video"
+        mock_asset.storage_key = "uploads/src.mp4"
+        mock_asset_repo = MagicMock()
+        mock_asset_repo.get.return_value = mock_asset
+
+        mock_mk = MagicMock()
+        mock_mk.is_available = True
+        mock_mk.extract_frames.return_value = [{"image_url": "https://mk/frame.jpg"}]
+
+        mock_storage = MagicMock()
+        mock_storage.get_url.return_value = "https://oss.example.com/uploads/src.mp4"
+
+        body = GenerateCoverRequest(cover_type="ai_frame", asset_ids=["a1"])
+
+        with (
+            patch("app.api.routes.generation_cover.SQLAlchemyGenerationTaskRepository") as mock_repo_cls,
+            patch(
+                "packages.adapters.sqlalchemy_impl.asset_repository.SQLAlchemyAssetRepository",
+                return_value=mock_asset_repo,
+            ),
+            patch("packages.shared.mediakit_client.get_mediakit_client", return_value=mock_mk),
+            patch("packages.shared.storage.get_shared_storage_service", return_value=mock_storage),
+            patch(
+                "app.api.routes.generation_cover._persist_cover_frame",
+                return_value="https://oss.example.com/covers/final.jpg",
+            ) as mock_persist,
+            patch("app.api.routes.generation_cover.normalize_plan_config") as mock_normalize,
+        ):
+            mock_repo = MagicMock()
+            mock_repo.get.return_value = None
+            mock_repo.list_by_source_edit_plan.return_value = []
+            mock_repo.list_latest_completed_preview.return_value = []
+            mock_repo_cls.return_value = mock_repo
+            mock_normalize.return_value = {
+                "cover": {"type": "ai_frame", "image_url": "https://oss.example.com/covers/final.jpg"}
+            }
+
+            from app.api.routes.generation_cover import generate_cover
+
+            result = generate_cover(
+                body=body,
+                template_id="tpl",
+                plan_id="plan-title",
+                services=(mock_template_svc, mock_plan_svc),
+                db=mock_db,
+                current_user=MagicMock(),
+            )
+
+            assert result.cover["image_url"] == "https://oss.example.com/covers/final.jpg"
+            # 标题文字必须透传给持久化函数（用于源素材帧叠加标题）
+            assert mock_persist.call_args.kwargs.get("title_text") == "我的视频标题"
+
+    def test_e2_passes_full_title_style_to_persist(self):
+        """步骤E2：plan.config.title 包含完整样式时，color/position/font_size 都传给 _persist_cover_frame。"""
+        from unittest.mock import MagicMock, patch
+
+        from app.api.routes.generation_cover import GenerateCoverRequest
+
+        mock_plan = MagicMock()
+        mock_plan.config = {
+            "title": {
+                "enabled": True,
+                "text": "样式标题",
+                "color": "#00ff00",
+                "position": "top",
+                "font_size": 42,
+            }
+        }
+
+        mock_plan_svc = MagicMock()
+        mock_plan_svc.get_plan_or_raise.return_value = mock_plan
+        mock_template_svc = MagicMock()
+        mock_db = MagicMock()
+
+        mock_asset = MagicMock()
+        mock_asset.file_type = "video"
+        mock_asset.storage_key = "uploads/src.mp4"
+        mock_asset_repo = MagicMock()
+        mock_asset_repo.get.return_value = mock_asset
+
+        mock_mk = MagicMock()
+        mock_mk.is_available = True
+        mock_mk.extract_frames.return_value = [{"image_url": "https://mk/frame.jpg"}]
+
+        mock_storage = MagicMock()
+        mock_storage.get_url.return_value = "https://oss.example.com/uploads/src.mp4"
+
+        body = GenerateCoverRequest(cover_type="ai_frame", asset_ids=["a1"])
+
+        with (
+            patch("app.api.routes.generation_cover.SQLAlchemyGenerationTaskRepository") as mock_repo_cls,
+            patch(
+                "packages.adapters.sqlalchemy_impl.asset_repository.SQLAlchemyAssetRepository",
+                return_value=mock_asset_repo,
+            ),
+            patch("packages.shared.mediakit_client.get_mediakit_client", return_value=mock_mk),
+            patch("packages.shared.storage.get_shared_storage_service", return_value=mock_storage),
+            patch(
+                "app.api.routes.generation_cover._persist_cover_frame",
+                return_value="https://oss.example.com/covers/styled.jpg",
+            ) as mock_persist,
+            patch("app.api.routes.generation_cover.normalize_plan_config") as mock_normalize,
+        ):
+            mock_repo = MagicMock()
+            mock_repo.get.return_value = None
+            mock_repo.list_by_source_edit_plan.return_value = []
+            mock_repo.list_latest_completed_preview.return_value = []
+            mock_repo_cls.return_value = mock_repo
+            mock_normalize.return_value = {
+                "cover": {"type": "ai_frame", "image_url": "https://oss.example.com/covers/styled.jpg"}
+            }
+
+            from app.api.routes.generation_cover import generate_cover
+
+            result = generate_cover(
+                body=body,
+                template_id="tpl",
+                plan_id="plan-style",
+                services=(mock_template_svc, mock_plan_svc),
+                db=mock_db,
+                current_user=MagicMock(),
+            )
+
+            assert result.cover["image_url"] == "https://oss.example.com/covers/styled.jpg"
+            kwargs = mock_persist.call_args.kwargs
+            assert kwargs["title_text"] == "样式标题"
+            assert kwargs["title_color"] == "#00ff00"
+            assert kwargs["title_position"] == "top"
+            assert kwargs["title_font_size"] == 42
+
+    def test_e2_title_style_fallback_font_color(self):
+        """步骤E2：前端传 font_color 时能正确兼容读取。"""
+        from unittest.mock import MagicMock, patch
+
+        from app.api.routes.generation_cover import GenerateCoverRequest
+
+        mock_plan = MagicMock()
+        mock_plan.config = {
+            "title": {
+                "enabled": True,
+                "text": "兼容标题",
+                "font_color": "#123456",
+                "position": "center",
+            }
+        }
+
+        mock_plan_svc = MagicMock()
+        mock_plan_svc.get_plan_or_raise.return_value = mock_plan
+        mock_template_svc = MagicMock()
+        mock_db = MagicMock()
+
+        mock_asset = MagicMock()
+        mock_asset.file_type = "video"
+        mock_asset.storage_key = "uploads/src.mp4"
+        mock_asset_repo = MagicMock()
+        mock_asset_repo.get.return_value = mock_asset
+
+        mock_mk = MagicMock()
+        mock_mk.is_available = True
+        mock_mk.extract_frames.return_value = [{"image_url": "https://mk/frame.jpg"}]
+
+        mock_storage = MagicMock()
+        mock_storage.get_url.return_value = "https://oss.example.com/uploads/src.mp4"
+
+        body = GenerateCoverRequest(cover_type="ai_frame", asset_ids=["a1"])
+
+        with (
+            patch("app.api.routes.generation_cover.SQLAlchemyGenerationTaskRepository") as mock_repo_cls,
+            patch(
+                "packages.adapters.sqlalchemy_impl.asset_repository.SQLAlchemyAssetRepository",
+                return_value=mock_asset_repo,
+            ),
+            patch("packages.shared.mediakit_client.get_mediakit_client", return_value=mock_mk),
+            patch("packages.shared.storage.get_shared_storage_service", return_value=mock_storage),
+            patch(
+                "app.api.routes.generation_cover._persist_cover_frame",
+                return_value="https://oss.example.com/covers/compat.jpg",
+            ) as mock_persist,
+            patch("app.api.routes.generation_cover.normalize_plan_config") as mock_normalize,
+        ):
+            mock_repo = MagicMock()
+            mock_repo.get.return_value = None
+            mock_repo.list_by_source_edit_plan.return_value = []
+            mock_repo.list_latest_completed_preview.return_value = []
+            mock_repo_cls.return_value = mock_repo
+            mock_normalize.return_value = {
+                "cover": {"type": "ai_frame", "image_url": "https://oss.example.com/covers/compat.jpg"}
+            }
+
+            from app.api.routes.generation_cover import generate_cover
+
+            generate_cover(
+                body=body,
+                template_id="tpl",
+                plan_id="plan-compat",
+                services=(mock_template_svc, mock_plan_svc),
+                db=mock_db,
+                current_user=MagicMock(),
+            )
+
+            kwargs = mock_persist.call_args.kwargs
+            assert kwargs["title_color"] == "#123456"
+            assert kwargs["title_position"] == "center"
+            assert kwargs["title_font_size"] is None
+
+    def test_step_e_skips_non_video_assets(self):
+        """步骤E2：asset_ids 里只有图片素材时，不调用 MediaKit 并返回 400。"""
+        from unittest.mock import MagicMock, patch
+
+        from app.api.routes.generation_cover import GenerateCoverRequest
+        from fastapi import HTTPException
+
+        mock_plan = MagicMock()
+        mock_plan.config = {}
+
+        mock_plan_svc = MagicMock()
+        mock_plan_svc.get_plan_or_raise.return_value = mock_plan
+        mock_template_svc = MagicMock()
+        mock_db = MagicMock()
+
+        mock_image_asset = MagicMock()
+        mock_image_asset.file_type = "image"
+        mock_image_asset.storage_key = "uploads/photo.png"
+
+        mock_asset_repo = MagicMock()
+        mock_asset_repo.get.return_value = mock_image_asset
+
+        mock_mk = MagicMock()
+        mock_mk.is_available = True
+
+        body = GenerateCoverRequest(
+            cover_type="ai_frame",
+            asset_ids=["asset-img-1"],
+        )
+
+        with (
+            patch("app.api.routes.generation_cover.SQLAlchemyGenerationTaskRepository") as mock_repo_cls,
+            patch(
+                "packages.adapters.sqlalchemy_impl.asset_repository.SQLAlchemyAssetRepository",
+                return_value=mock_asset_repo,
+            ),
+            patch("packages.shared.mediakit_client.get_mediakit_client", return_value=mock_mk),
+            patch("packages.shared.storage.get_shared_storage_service") as mock_storage_getter,
+        ):
+            mock_repo = MagicMock()
+            mock_repo.get.return_value = None
+            mock_repo.list_by_source_edit_plan.return_value = []
+            mock_repo.list_latest_completed_preview.return_value = []
+            mock_repo_cls.return_value = mock_repo
+            mock_storage_getter.return_value = MagicMock()
+
+            from app.api.routes.generation_cover import generate_cover
+
+            with pytest.raises(HTTPException) as exc_info:
+                generate_cover(
+                    body=body,
+                    template_id="tpl-img",
+                    plan_id="plan-img",
+                    services=(mock_template_svc, mock_plan_svc),
+                    db=mock_db,
+                    current_user=MagicMock(),
+                )
+
+            assert exc_info.value.status_code == 400
+            mock_mk.extract_frames.assert_not_called()

tests/unit/test_generation_task_pending_cleanup.py

@@ -0,0 +1,151 @@
+"""GenerationTaskRepository - cleanup_stale_pending 超时 pending 清理单元测试。"""
+
+import sys
+from datetime import datetime, timedelta, timezone
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[2] / "apps" / "api"))
+
+from sqlalchemy import create_engine, text
+from sqlalchemy.orm import sessionmaker
+
+from packages.adapters.sqlalchemy_impl.generation_task_repository import (
+    SQLAlchemyGenerationTaskRepository,
+)
+from packages.adapters.sqlalchemy_impl.models import Base
+from packages.domain import GenerationTask, GenerationTaskStatus
+
+
+def _repository():
+    engine = create_engine("sqlite:///:memory:")
+    Base.metadata.create_all(engine)
+    session = sessionmaker(bind=engine)()
+    return SQLAlchemyGenerationTaskRepository(session), session, engine
+
+
+def _make_task(**kwargs) -> GenerationTask:
+    defaults = dict(
+        project_id="proj-1",
+        asset_library_id="lib-1",
+        created_by_user_id="user-1",
+    )
+    defaults.update(kwargs)
+    return GenerationTask.create(**defaults)
+
+
+# ---------------------------------------------------------------------------
+# cleanup_stale_pending 基本测试
+# ---------------------------------------------------------------------------
+
+
+def test_cleanup_stale_pending_no_tasks_returns_zero():
+    """没有任务时返回 0。"""
+    repo, _, _ = _repository()
+    count = repo.cleanup_stale_pending(timeout_minutes=30)
+    assert count == 0
+
+
+def test_cleanup_stale_pending_recent_pending_not_cleaned():
+    """30 分钟内的 pending 任务不被清理。"""
+    repo, _, _ = _repository()
+    task = _make_task()
+    repo.create(task)
+    # 刚创建的 pending 任务不应被清理
+    count = repo.cleanup_stale_pending(timeout_minutes=30)
+    assert count == 0
+    assert repo.get(task.id).status == GenerationTaskStatus.PENDING
+
+
+def test_cleanup_stale_pending_old_pending_marked_failed():
+    """超过 30 分钟的 pending 任务被标记为 failed。"""
+    repo, _, engine = _repository()
+    task = _make_task()
+    repo.create(task)
+
+    # 手动把 created_at 改到 1 小时前
+    with engine.connect() as conn:
+        conn.execute(
+            text("UPDATE generation_tasks SET created_at = :ts WHERE id = :id"),
+            {"ts": datetime.now(timezone.utc) - timedelta(hours=1), "id": task.id},
+        )
+        conn.commit()
+
+    count = repo.cleanup_stale_pending(timeout_minutes=30)
+    assert count == 1
+
+    saved = repo.get(task.id)
+    assert saved.status == GenerationTaskStatus.FAILED
+    assert saved.error_message == "pending timeout: auto cleanup"
+    assert saved.error_info.get("error_type") == "PendingTimeout"
+    assert "30" in saved.error_info["message"]
+    assert "failed_at" in saved.error_info
+    assert saved.completed_at is not None
+
+
+def test_cleanup_stale_pending_running_not_touched():
+    """running 任务不受影响，只清理 pending。"""
+    repo, _, engine = _repository()
+    task = _make_task()
+    repo.create(task)
+    task.mark_processing()
+    repo.update(task)
+
+    # 回写 created_at 到 1 小时前
+    with engine.connect() as conn:
+        conn.execute(
+            text("UPDATE generation_tasks SET created_at = :ts WHERE id = :id"),
+            {"ts": datetime.now(timezone.utc) - timedelta(hours=1), "id": task.id},
+        )
+        conn.commit()
+
+    count = repo.cleanup_stale_pending(timeout_minutes=30)
+    assert count == 0
+    assert repo.get(task.id).status == GenerationTaskStatus.RUNNING
+
+
+def test_cleanup_stale_pending_custom_timeout():
+    """自定义超时时间生效。"""
+    repo, _, engine = _repository()
+    task = _make_task()
+    repo.create(task)
+
+    # 回写 created_at 到 20 分钟前
+    with engine.connect() as conn:
+        conn.execute(
+            text("UPDATE generation_tasks SET created_at = :ts WHERE id = :id"),
+            {"ts": datetime.now(timezone.utc) - timedelta(minutes=20), "id": task.id},
+        )
+        conn.commit()
+
+    # 30 分钟超时：不清理
+    count_30 = repo.cleanup_stale_pending(timeout_minutes=30)
+    assert count_30 == 0
+    # 15 分钟超时：清理
+    count_15 = repo.cleanup_stale_pending(timeout_minutes=15)
+    assert count_15 == 1
+    assert repo.get(task.id).status == GenerationTaskStatus.FAILED
+
+
+def test_cleanup_stale_pending_multiple():
+    """批量清理多个超时的 pending 任务。"""
+    repo, _, engine = _repository()
+
+    tasks = []
+    for i in range(5):
+        t = _make_task(project_id=f"proj-{i}")
+        repo.create(t)
+        tasks.append(t)
+
+    # 全部回写 created_at 到 2 小时前
+    with engine.connect() as conn:
+        for t in tasks:
+            conn.execute(
+                text("UPDATE generation_tasks SET created_at = :ts WHERE id = :id"),
+                {"ts": datetime.now(timezone.utc) - timedelta(hours=2), "id": t.id},
+            )
+        conn.commit()
+
+    count = repo.cleanup_stale_pending(timeout_minutes=30)
+    assert count == 5
+    for t in tasks:
+        assert repo.get(t.id).status == GenerationTaskStatus.FAILED

tests/unit/test_generation_worker_fixes.py

@@ -13,6 +13,18 @@ os.environ.setdefault("JWT_SECRET_KEY", "unit-test-secret")
 os.environ.setdefault("DATABASE_URL", "sqlite:///test.db")
 
 
+def _patch_session_local(mock_session):
+    """Patch worker_app.db.SessionLocal robustly even when other tests
+    have pre-registered a MagicMock for worker_app.db in sys.modules.
+    Uses patch.dict to inject a clean module so that
+    'from worker_app.db import SessionLocal' resolves correctly."""
+    from types import ModuleType
+
+    _fresh_db = ModuleType("worker_app.db")
+    _fresh_db.SessionLocal = lambda *a, **kw: mock_session
+    return patch.dict(sys.modules, {"worker_app.db": _fresh_db})
+
+
 class TestLoadTemplateSegmentDurations:
     """_load_template_segment_durations 单元测试 (covers lines 198-226)."""
 
@@ -40,8 +52,7 @@ class TestLoadTemplateSegmentDurations:
         mock_session = MagicMock()
         mock_session.query.return_value = mock_query
 
-        # Patch at the source module since it's imported inside the function
-        with patch("worker_app.db.SessionLocal", return_value=mock_session):
+        with _patch_session_local(mock_session):
             result = _load_template_segment_durations("tpl_123")
 
         assert result == [5.0, 8.0, 3.0]
@@ -63,16 +74,24 @@ class TestLoadTemplateSegmentDurations:
         mock_session = MagicMock()
         mock_session.query.return_value = mock_query
 
-        with patch("worker_app.db.SessionLocal", return_value=mock_session):
+        with _patch_session_local(mock_session):
             result = _load_template_segment_durations("tpl_456")
 
         assert result == [5.0]
 
     def test_db_error_returns_empty(self):
         """数据库异常返回空列表，不抛出。"""
+        from types import ModuleType
+
         from worker_app.tasks.generation import _load_template_segment_durations
 
-        with patch("worker_app.db.SessionLocal", side_effect=Exception("DB down")):
+        _err_db = ModuleType("worker_app.db")
+
+        def _raise(*a, **kw):
+            raise Exception("DB down")
+
+        _err_db.SessionLocal = _raise
+        with patch.dict(sys.modules, {"worker_app.db": _err_db}):
             result = _load_template_segment_durations("tpl_789")
 
         assert result == []
@@ -86,7 +105,7 @@ class TestLoadTemplateSegmentDurations:
         mock_session = MagicMock()
         mock_session.query.return_value = mock_query
 
-        with patch("worker_app.db.SessionLocal", return_value=mock_session):
+        with _patch_session_local(mock_session):
             result = _load_template_segment_durations("tpl_empty")
 
         assert result == []

tests/unit/test_health_routes.py

@@ -1,7 +1,6 @@
 """Unit tests for apps/api/app/api/routes/health.py
 
 覆盖 _check_database() 和 _check_migrations() 中 psycopg3 连接逻辑。
-确保增量覆盖率 ≥ 60%（目标覆盖 lines 52, 127）。
 """
 
 from unittest.mock import AsyncMock, MagicMock, patch
@@ -9,63 +8,70 @@ from unittest.mock import AsyncMock, MagicMock, patch
 import pytest
 
 
+def _make_cursor(fetchone_result=None):
+    """Create a mock cursor with context manager support."""
+    cur = MagicMock()
+    cur.__enter__ = MagicMock(return_value=cur)
+    cur.__exit__ = MagicMock(return_value=False)
+    if fetchone_result is not None:
+        cur.fetchone.return_value = fetchone_result
+    return cur
+
+
+def _make_conn(cursor_result=None):
+    conn = MagicMock()
+    conn.cursor.return_value = cursor_result or _make_cursor()
+    return conn
+
+
 @pytest.mark.asyncio
 class TestCheckDatabase:
-    """Tests for _check_database() health check function."""
 
-    @patch("apps.api.app.api.routes.health.settings")
-    @patch("apps.api.app.api.routes.health.psycopg.connect")
-    async def test_check_database_success(self, mock_connect, mock_settings):
-        """PostgreSQL 连接成功时返回 healthy。"""
+    async def test_check_database_success(self):
+        mock_cur = _make_cursor(fetchone_result=(1,))
+        mock_conn = _make_conn(mock_cur)
+        mock_settings = MagicMock()
         mock_settings.USE_IN_MEMORY_DB = False
         mock_settings.DATABASE_URL = "postgresql+psycopg://test:test@localhost/test"
 
-        # Mock connection and cursor
-        mock_conn = MagicMock()
-        mock_cursor = MagicMock()
-        mock_cursor.__enter__ = MagicMock(return_value=mock_cursor)
-        mock_cursor.__exit__ = MagicMock(return_value=False)
-        mock_cursor.fetchone.return_value = (1,)
-        mock_conn.cursor.return_value = mock_cursor
-        mock_connect.return_value = mock_conn
+        from apps.api.app.api.routes import health
 
-        from apps.api.app.api.routes.health import _check_database
-
-        result = await _check_database()
+        with patch.object(health, "psycopg") as mock_psycopg,              patch.object(health, "settings", mock_settings):
+            mock_psycopg.connect.return_value = mock_conn
+            result = await health._check_database()
 
         assert result["status"] == "healthy"
         assert result["type"] == "postgresql"
         assert result["message"] == "Database connection successful"
-        mock_connect.assert_called_once_with(
+        mock_psycopg.connect.assert_called_once_with(
             "postgresql+psycopg://test:test@localhost/test", connect_timeout=3
         )
-        mock_cursor.execute.assert_called_once_with("SELECT 1")
+        mock_cur.execute.assert_called_once_with("SELECT 1")
         mock_conn.close.assert_called_once()
 
-    @patch("apps.api.app.api.routes.health.settings")
-    @patch("apps.api.app.api.routes.health.psycopg.connect")
-    async def test_check_database_connection_failure(self, mock_connect, mock_settings):
-        """PostgreSQL 连接失败时返回 unhealthy。"""
+    async def test_check_database_connection_failure(self):
+        mock_settings = MagicMock()
         mock_settings.USE_IN_MEMORY_DB = False
         mock_settings.DATABASE_URL = "postgresql+psycopg://test:test@localhost/test"
-        mock_connect.side_effect = Exception("connection refused")
 
-        from apps.api.app.api.routes.health import _check_database
+        from apps.api.app.api.routes import health
 
-        result = await _check_database()
+        with patch.object(health, "psycopg") as mock_psycopg,              patch.object(health, "settings", mock_settings):
+            mock_psycopg.connect.side_effect = Exception("connection refused")
+            result = await health._check_database()
 
         assert result["status"] == "unhealthy"
         assert result["type"] == "postgresql"
         assert "connection refused" in result["message"]
 
-    @patch("apps.api.app.api.routes.health.settings")
-    async def test_check_database_in_memory(self, mock_settings):
-        """使用内存数据库时跳过 PostgreSQL 检查。"""
+    async def test_check_database_in_memory(self):
+        mock_settings = MagicMock()
         mock_settings.USE_IN_MEMORY_DB = True
 
-        from apps.api.app.api.routes.health import _check_database
+        from apps.api.app.api.routes import health
 
-        result = await _check_database()
+        with patch.object(health, "settings", mock_settings):
+            result = await health._check_database()
 
         assert result["status"] == "healthy"
         assert result["type"] == "in_memory"
@@ -73,80 +79,66 @@ class TestCheckDatabase:
 
 @pytest.mark.asyncio
 class TestCheckMigrations:
-    """Tests for _check_migrations() health check function."""
 
-    @patch("apps.api.app.api.routes.health.settings")
-    @patch("apps.api.app.api.routes.health.psycopg.connect")
-    async def test_check_migrations_success(self, mock_connect, mock_settings):
-        """所有迁移表存在时返回 healthy。"""
+    async def test_check_migrations_success(self):
+        mock_cur = _make_cursor(fetchone_result=(5,))
+        mock_conn = _make_conn(mock_cur)
+        mock_settings = MagicMock()
         mock_settings.USE_IN_MEMORY_DB = False
         mock_settings.DATABASE_URL = "postgresql+psycopg://test:test@localhost/test"
 
-        mock_conn = MagicMock()
-        mock_cursor = MagicMock()
-        mock_cursor.__enter__ = MagicMock(return_value=mock_cursor)
-        mock_cursor.__exit__ = MagicMock(return_value=False)
-        mock_cursor.fetchone.return_value = (5,)  # 5 tables found
-        mock_conn.cursor.return_value = mock_cursor
-        mock_connect.return_value = mock_conn
+        from apps.api.app.api.routes import health
 
-        from apps.api.app.api.routes.health import _check_migrations
-
-        result = await _check_migrations()
+        with patch.object(health, "psycopg") as mock_psycopg,              patch.object(health, "settings", mock_settings):
+            mock_psycopg.connect.return_value = mock_conn
+            result = await health._check_migrations()
 
         assert result["status"] == "healthy"
         assert result["message"] == "Database migrations applied"
-        mock_connect.assert_called_once_with(
+        mock_psycopg.connect.assert_called_once_with(
             "postgresql+psycopg://test:test@localhost/test", connect_timeout=3
         )
         mock_conn.close.assert_called_once()
 
-    @patch("apps.api.app.api.routes.health.settings")
-    @patch("apps.api.app.api.routes.health.psycopg.connect")
-    async def test_check_migrations_missing_tables(self, mock_connect, mock_settings):
-        """迁移表不完整时返回 unhealthy。"""
+    async def test_check_migrations_missing_tables(self):
+        mock_cur = _make_cursor(fetchone_result=(2,))
+        mock_conn = _make_conn(mock_cur)
+        mock_settings = MagicMock()
         mock_settings.USE_IN_MEMORY_DB = False
         mock_settings.DATABASE_URL = "postgresql+psycopg://test:test@localhost/test"
 
-        mock_conn = MagicMock()
-        mock_cursor = MagicMock()
-        mock_cursor.__enter__ = MagicMock(return_value=mock_cursor)
-        mock_cursor.__exit__ = MagicMock(return_value=False)
-        mock_cursor.fetchone.return_value = (2,)  # Only 2 of 5 tables
-        mock_conn.cursor.return_value = mock_cursor
-        mock_connect.return_value = mock_conn
+        from apps.api.app.api.routes import health
 
-        from apps.api.app.api.routes.health import _check_migrations
-
-        result = await _check_migrations()
+        with patch.object(health, "psycopg") as mock_psycopg,              patch.object(health, "settings", mock_settings):
+            mock_psycopg.connect.return_value = mock_conn
+            result = await health._check_migrations()
 
         assert result["status"] == "unhealthy"
         assert "Missing tables" in result["message"]
         assert "2/5" in result["message"]
 
-    @patch("apps.api.app.api.routes.health.settings")
-    @patch("apps.api.app.api.routes.health.psycopg.connect")
-    async def test_check_migrations_connection_failure(self, mock_connect, mock_settings):
-        """数据库连接失败时返回 unhealthy。"""
+    async def test_check_migrations_connection_failure(self):
+        mock_settings = MagicMock()
         mock_settings.USE_IN_MEMORY_DB = False
         mock_settings.DATABASE_URL = "postgresql+psycopg://test:test@localhost/test"
-        mock_connect.side_effect = Exception("connection refused")
 
-        from apps.api.app.api.routes.health import _check_migrations
+        from apps.api.app.api.routes import health
 
-        result = await _check_migrations()
+        with patch.object(health, "psycopg") as mock_psycopg,              patch.object(health, "settings", mock_settings):
+            mock_psycopg.connect.side_effect = Exception("connection refused")
+            result = await health._check_migrations()
 
         assert result["status"] == "unhealthy"
         assert "Migration check failed" in result["message"]
 
-    @patch("apps.api.app.api.routes.health.settings")
-    async def test_check_migrations_in_memory(self, mock_settings):
-        """使用内存数据库时跳过迁移检查。"""
+    async def test_check_migrations_in_memory(self):
+        mock_settings = MagicMock()
         mock_settings.USE_IN_MEMORY_DB = True
 
-        from apps.api.app.api.routes.health import _check_migrations
+        from apps.api.app.api.routes import health
 
-        result = await _check_migrations()
+        with patch.object(health, "settings", mock_settings):
+            result = await health._check_migrations()
 
         assert result["status"] == "healthy"
         assert "no migrations needed" in result["message"]
@@ -154,33 +146,27 @@ class TestCheckMigrations:
 
 @pytest.mark.asyncio
 class TestStartupCheck:
-    """Tests for startup_check() endpoint."""
 
-    @patch("apps.api.app.api.routes.health._check_migrations")
-    @patch("apps.api.app.api.routes.health._check_database")
-    async def test_startup_all_healthy(self, mock_db, mock_mig):
-        """所有检查通过时返回 started。"""
-        mock_db.return_value = {"status": "healthy"}
-        mock_mig.return_value = {"status": "healthy"}
+    async def test_startup_all_healthy(self):
+        from apps.api.app.api.routes import health
 
-        from apps.api.app.api.routes.health import startup_check
-
-        result = await startup_check()
+        with patch.object(health, "_check_migrations", new_callable=AsyncMock) as mock_mig,              patch.object(health, "_check_database", new_callable=AsyncMock) as mock_db:
+            mock_db.return_value = {"status": "healthy"}
+            mock_mig.return_value = {"status": "healthy"}
+            result = await health.startup_check()
 
         assert result["status"] == "started"
 
-    @patch("apps.api.app.api.routes.health._check_migrations")
-    @patch("apps.api.app.api.routes.health._check_database")
-    async def test_startup_db_unhealthy(self, mock_db, mock_mig):
-        """数据库不健康时返回 starting + 503。"""
-        mock_db.return_value = {"status": "unhealthy", "message": "fail"}
-        mock_mig.return_value = {"status": "healthy"}
+    async def test_startup_db_unhealthy(self):
+        import json
 
-        from apps.api.app.api.routes.health import startup_check
+        from apps.api.app.api.routes import health
 
-        result = await startup_check()
+        with patch.object(health, "_check_migrations", new_callable=AsyncMock) as mock_mig,              patch.object(health, "_check_database", new_callable=AsyncMock) as mock_db:
+            mock_db.return_value = {"status": "unhealthy", "message": "fail"}
+            mock_mig.return_value = {"status": "healthy"}
+            result = await health.startup_check()
 
         assert result.status_code == 503
-        import json
         body = json.loads(result.body)
         assert body["status"] == "starting"

tests/unit/test_replace_all_clips_transactional.py

@@ -0,0 +1,147 @@
+"""Tests for EditPlanService.replace_all_clips_transactional."""
+
+from __future__ import annotations
+
+import os
+from unittest.mock import MagicMock, PropertyMock, patch
+
+import pytest
+
+os.environ.setdefault("JWT_SECRET_KEY", "unit-test-secret-key-for-testing")
+
+
+class TestReplaceAllClipsTransactional:
+    """事务性替换片段方法测试。"""
+
+    @patch("packages.adapters.sqlalchemy_impl.models.EditPlanClipModel")
+    @patch("app.services.edit_plan_service.EditPlanClip")
+    def test_success_commits_once(self, mock_clip_cls, mock_model_cls):
+        """成功时单次 commit，不 rollback。"""
+        from app.services.edit_plan_service import EditPlanService
+
+        db = MagicMock()
+        # Mock query chain for delete
+        query_mock = MagicMock()
+        query_mock.filter.return_value.delete.return_value = 3
+        db.query.return_value = query_mock
+
+        # Mock query chain for mark_ready (pending_with_asset)
+        # After the create loop, query returns empty list (no pending clips with asset)
+        ready_query = MagicMock()
+        ready_query.filter.return_value.filter.return_value.filter.return_value.all.return_value = []
+        db.query.side_effect = [query_mock, ready_query]
+
+        # Mock EditPlanClip.create to return a mock entity
+        mock_entity = MagicMock()
+        mock_entity.id = "clip-1"
+        mock_entity.plan_id = "plan-1"
+        mock_entity.clip_type = "main"
+        mock_entity.order = 0
+        mock_entity.asset_id = "asset-1"
+        mock_entity.text_content = ""
+        mock_entity.start_time = 0.0
+        mock_entity.duration = 3.0
+        mock_entity.transition_effect = "cut"
+        mock_entity.transition_duration = 0.0
+        mock_entity.playback_speed = 1.0
+        mock_entity.status.value = "pending"
+        mock_entity.config = {}
+        mock_clip_cls.create.return_value = mock_entity
+
+        # Mock the model constructor
+        mock_model_instance = MagicMock()
+        mock_model_cls.return_value = mock_model_instance
+
+        # Mock clip_repo
+        clip_repo = MagicMock()
+        clip_repo.session = db
+
+        svc = EditPlanService.__new__(EditPlanService)
+        svc._clip_repo = clip_repo
+
+        result = svc.replace_all_clips_transactional(
+            "plan-1",
+            [{"asset_id": "asset-1", "start_time": 0.0, "duration": 3.0, "order": 0}],
+        )
+
+        assert result == 1
+        db.commit.assert_called_once()
+        db.rollback.assert_not_called()
+        db.add.assert_called_once_with(mock_model_instance)
+
+    @patch("packages.adapters.sqlalchemy_impl.models.EditPlanClipModel")
+    @patch("app.services.edit_plan_service.EditPlanClip")
+    def test_failure_rolls_back(self, mock_clip_cls, mock_model_cls):
+        """异常时自动 rollback。"""
+        from app.services.edit_plan_service import EditPlanService
+
+        db = MagicMock()
+        query_mock = MagicMock()
+        query_mock.filter.return_value.delete.return_value = 0
+        db.query.return_value = query_mock
+
+        # Simulate failure during create
+        mock_clip_cls.create.side_effect = ValueError("模拟异常")
+
+        clip_repo = MagicMock()
+        clip_repo.session = db
+
+        svc = EditPlanService.__new__(EditPlanService)
+        svc._clip_repo = clip_repo
+
+        with pytest.raises(ValueError, match="模拟异常"):
+            svc.replace_all_clips_transactional(
+                "plan-1",
+                [{"asset_id": "bad", "start_time": 0.0, "duration": 1.0, "order": 0}],
+            )
+
+        db.rollback.assert_called_once()
+        db.commit.assert_not_called()
+
+    @patch("packages.adapters.sqlalchemy_impl.models.EditPlanClipModel")
+    @patch("app.services.edit_plan_service.EditPlanClip")
+    def test_order_defaults_to_index(self, mock_clip_cls, mock_model_cls):
+        """order=0 时使用索引值作为 order。"""
+        from app.services.edit_plan_service import EditPlanService
+
+        db = MagicMock()
+        query_mock = MagicMock()
+        query_mock.filter.return_value.delete.return_value = 0
+        db.query.return_value = query_mock
+
+        ready_query = MagicMock()
+        ready_query.filter.return_value.filter.return_value.filter.return_value.all.return_value = []
+        db.query.side_effect = [query_mock, ready_query]
+
+        mock_entity = MagicMock()
+        mock_entity.id = "clip-1"
+        mock_entity.plan_id = "plan-1"
+        mock_entity.clip_type = "main"
+        mock_entity.order = 0  # order=0 → 使用 i=0
+        mock_entity.asset_id = "a1"
+        mock_entity.text_content = ""
+        mock_entity.start_time = 0.0
+        mock_entity.duration = 1.0
+        mock_entity.transition_effect = "cut"
+        mock_entity.transition_duration = 0.0
+        mock_entity.playback_speed = 1.0
+        mock_entity.status.value = "pending"
+        mock_entity.config = {}
+        mock_clip_cls.create.return_value = mock_entity
+
+        mock_model_cls.return_value = MagicMock()
+
+        clip_repo = MagicMock()
+        clip_repo.session = db
+
+        svc = EditPlanService.__new__(EditPlanService)
+        svc._clip_repo = clip_repo
+
+        svc.replace_all_clips_transactional(
+            "plan-1",
+            [{"asset_id": "a1", "start_time": 0.0, "duration": 1.0, "order": 0}],
+        )
+
+        # order=0 → falsy → use index i=0
+        create_call = mock_clip_cls.create.call_args
+        assert create_call.kwargs["order"] == 0

tests/unit/test_title_overlay.py

@@ -0,0 +1,89 @@
+"""packages.shared.title_overlay 单元测试。"""
+
+from __future__ import annotations
+
+import tempfile
+from pathlib import Path
+
+import pytest
+
+from packages.shared.title_overlay import apply_title_to_image, wrap_title_text
+
+
+@pytest.fixture
+def sample_image():
+    """生成一张 640x360 的纯黑测试图片。"""
+    from PIL import Image
+
+    tmp = tempfile.NamedTemporaryFile(suffix=".jpg", delete=False)
+    tmp.close()
+    img = Image.new("RGB", (640, 360), color=(0, 0, 0))
+    img.save(tmp.name, "JPEG")
+    yield tmp.name
+    Path(tmp.name).unlink(missing_ok=True)
+
+
+def test_apply_title_to_image_empty_text_returns_none(sample_image):
+    assert apply_title_to_image(sample_image, "") is None
+    assert apply_title_to_image(sample_image, "   ") is None
+
+
+def test_apply_title_to_image_draws_title(sample_image):
+    result = apply_title_to_image(sample_image, "测试标题")
+    assert result == sample_image
+    assert Path(sample_image).exists()
+    assert Path(sample_image).stat().st_size > 0
+
+
+def test_apply_title_to_image_respects_position(sample_image):
+    for pos in ("top", "center", "bottom"):
+        result = apply_title_to_image(sample_image, "位置测试", position=pos)
+        assert result == sample_image
+
+
+def test_wrap_title_text_supports_long_text():
+    from PIL import ImageFont
+
+    font = ImageFont.load_default()
+    lines = wrap_title_text("这是一个比较长的标题需要自动换行处理ABC", font, max_width=40)
+    assert isinstance(lines, list)
+    assert len(lines) >= 1
+
+
+def test_wrap_title_text_respects_explicit_newline():
+    from PIL import ImageFont
+
+    font = ImageFont.load_default()
+    lines = wrap_title_text("第一行\n第二行", font, max_width=10000)
+    assert lines == ["第一行", "第二行"]
+
+
+def test_apply_title_to_image_custom_color(sample_image):
+    """自定义颜色参数能正常生成图片。"""
+    result = apply_title_to_image(sample_image, "彩色标题", color="#ff0000")
+    assert result == sample_image
+    assert Path(sample_image).stat().st_size > 0
+
+
+def test_apply_title_to_image_short_hex_color(sample_image):
+    """3 位缩写 hex 颜色也能正常解析。"""
+    result = apply_title_to_image(sample_image, "短色", color="#f00")
+    assert result == sample_image
+
+
+def test_apply_title_to_image_invalid_color_fallback(sample_image):
+    """无效颜色字符串 fallback 到白色，不报错。"""
+    result = apply_title_to_image(sample_image, "异常色", color="not-a-color")
+    assert result == sample_image
+
+
+def test_parse_hex_color():
+    from packages.shared.title_overlay import _parse_hex_color
+
+    assert _parse_hex_color("#ffffff") == (255, 255, 255)
+    assert _parse_hex_color("#000000") == (0, 0, 0)
+    assert _parse_hex_color("#ff0000") == (255, 0, 0)
+    assert _parse_hex_color("#f00") == (255, 0, 0)
+    assert _parse_hex_color("") == (255, 255, 255)
+    assert _parse_hex_color("invalid") == (255, 255, 255)
+    assert _parse_hex_color("#gggggg") == (255, 255, 255)

tests/unit/test_writeback_edit_plan_config.py

@@ -0,0 +1,126 @@
+"""Tests for _writeback_edit_plan_config in generation_tasks route.
+
+覆盖 CI 增量覆盖率不足的代码：
+- generation_tasks.py 行 160-193 (_writeback_edit_plan_config 函数体)
+- generation_tasks.py 行 371-372 (路由中调用该函数)
+"""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[2] / "apps" / "api"))
+
+from app.api.routes.generation_tasks import _writeback_edit_plan_config
+
+
+@pytest.fixture
+def mock_db():
+    """Mock SQLAlchemy Session."""
+    db = MagicMock()
+    db.query.return_value = db
+    db.filter.return_value = db
+    return db
+
+
+@pytest.fixture
+def mock_plan():
+    """Mock EditPlanModel instance."""
+    plan = MagicMock()
+    plan.config = {"existing_key": "existing_value"}
+    return plan
+
+
+class TestWritebackEditPlanConfig:
+    """_writeback_edit_plan_config 全分支覆盖"""
+
+    # ---- 行 160-161: plan_id 为空直接返回 ----
+    def test_empty_plan_id_returns_immediately(self, mock_db):
+        _writeback_edit_plan_config(plan_id="", task_id="task_1", title_config={"text": "hi"}, db=mock_db)
+        mock_db.query.assert_not_called()
+        mock_db.commit.assert_not_called()
+
+    def test_none_plan_id_returns_immediately(self, mock_db):
+        _writeback_edit_plan_config(plan_id=None, task_id="task_1", title_config=None, db=mock_db)
+        mock_db.query.assert_not_called()
+
+    # ---- 行 165-168: plan 不存在 → warning + 不 commit ----
+    def test_plan_not_found_no_commit(self, mock_db):
+        mock_db.first.return_value = None
+
+        _writeback_edit_plan_config(plan_id="plan_999", task_id="task_1", title_config=None, db=mock_db)
+
+        mock_db.query.assert_called_once()
+        mock_db.commit.assert_not_called()
+
+    # ---- 行 170-182: 正常写入 + title_config ----
+    def test_success_with_title_config(self, mock_db, mock_plan):
+        mock_db.first.return_value = mock_plan
+
+        _writeback_edit_plan_config(
+            plan_id="plan_123",
+            task_id="task_456",
+            title_config={"text": "标题", "font_size": 36},
+            db=mock_db,
+        )
+
+        assert mock_plan.config["generation_task_id"] == "task_456"
+        assert mock_plan.config["title_config"] == {"text": "标题", "font_size": 36}
+        assert mock_plan.config["existing_key"] == "existing_value"
+        mock_db.commit.assert_called_once()
+
+    # ---- 行 170-175: 正常写入、无 title_config ----
+    def test_success_without_title_config(self, mock_db, mock_plan):
+        mock_db.first.return_value = mock_plan
+
+        _writeback_edit_plan_config(plan_id="plan_123", task_id="task_789", title_config=None, db=mock_db)
+
+        assert mock_plan.config["generation_task_id"] == "task_789"
+        assert "title_config" not in mock_plan.config
+        mock_db.commit.assert_called_once()
+
+    # ---- 行 170: config 不是 dict → 兜底空 dict ----
+    def test_config_not_dict_uses_empty_dict(self, mock_db):
+        bad_plan = MagicMock()
+        bad_plan.config = "not_a_dict"
+        mock_db.first.return_value = bad_plan
+
+        _writeback_edit_plan_config(plan_id="plan_123", task_id="task_1", title_config=None, db=mock_db)
+
+        assert isinstance(bad_plan.config, dict)
+        assert bad_plan.config["generation_task_id"] == "task_1"
+        mock_db.commit.assert_called_once()
+
+    # ---- 行 183-189: DB 异常 → warning + rollback ----
+    def test_db_exception_triggers_rollback(self, mock_db, mock_plan):
+        mock_db.first.return_value = mock_plan
+        mock_db.commit.side_effect = RuntimeError("DB connection lost")
+
+        # 不应抛异常
+        _writeback_edit_plan_config(plan_id="plan_123", task_id="task_1", title_config=None, db=mock_db)
+
+        mock_db.rollback.assert_called_once()
+
+    # ---- 行 190-193: rollback 也失败 → 静默 ----
+    def test_rollback_failure_silent(self, mock_db, mock_plan):
+        mock_db.first.return_value = mock_plan
+        mock_db.commit.side_effect = RuntimeError("commit failed")
+        mock_db.rollback.side_effect = RuntimeError("rollback also failed")
+
+        # 两个异常都不应抛出
+        _writeback_edit_plan_config(plan_id="plan_123", task_id="task_1", title_config=None, db=mock_db)
+        mock_db.rollback.assert_called_once()
+
+    # ---- 行 173: title_config 为空 dict → 不写入 title_config ----
+    def test_empty_title_config_not_written(self, mock_db, mock_plan):
+        mock_db.first.return_value = mock_plan
+
+        _writeback_edit_plan_config(plan_id="plan_123", task_id="task_1", title_config={}, db=mock_db)
+
+        # 空 dict 为 falsy，不写入
+        assert "title_config" not in mock_plan.config
+        assert mock_plan.config["generation_task_id"] == "task_1"