3602 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
CI Bot	fdad465d3a	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-20 14:36:48 +00:00
saas-backend	1584a85bab	feat(#2001): 爆款标题样式面板后端支持 ... - SubtitleStyle 新增 line_overrides 字段，支持逐行独立样式覆盖 (color/font_size/font/bold/italic/stroke_color/stroke_width/shadow) 通过 ASS 内联 override 标签 {\c&H...\fsN\fn...} 实现 - ass_subtitle_builder 在标题/字幕换行后应用 line_overrides - AiAvatarRenderJob 新增 cover_title_config 字段（封面独立标题配置，结构同 title_config） - Alembic 迁移 083: ai_avatar_render_jobs.cover_title_config JSON NOT NULL 默认 {} - Schema/Service/Route 全链路透传 cover_title_config - 部署 4 款爆款字体到 /usr/share/fonts/truetype/xiaoxia/: 优设标题黑、阿里普惠体 Bold 2.0、抖音美好体、思源黑体 Heavy - FONT_NAME_MAP + DRAWTEXT_FONT_MAP + DRAWTEXT_FONT_SEARCH_PATHS 全量扩展 - 单元测试 24 个覆盖 line_overrides 逐行标签/负数索引/全字段组合/异常过滤 - 覆盖 #1263 标题预设 T 图标能力（通过 line_overrides 首行大字号+粗体实现）	2026-09-20 21:57:54 +08:00
xiaoxia	dd7a5ad658	fix: 增强MuseTalk视觉质量 - GFPGAN面部增强 + 帧率适配 + 高质量音频重采样 ... P0: 集成GFPGAN面部增强后处理 - 懒加载GFPGAN模型，首次推理时自动加载并复用 - 对嘴部/面部区域做增强，提升锐度和纹理细节 - 支持mouth/face两种增强区域模式 - 高斯模糊遮罩平滑融合，消除接缝 P1: 60fps视频帧率适配 - 检测输入视频帧率，>30fps时自动降至25fps推理 - 推理完成后使用minterpolate运动补偿插帧恢复原始帧率 - 消除MuseTalk训练帧率不匹配导致的重复帧问题 P2: 高质量音频重采样 - 使用ffmpeg soxr重采样器(精度28)替代默认线性插值 - 不可用时自动回退默认重采样器 P1: 融合参数优化 - v1.5模型扩大bbox融合区域(下巴+15px，两侧各+2px) - 减少AI生成区域与原画面的接缝感 /inference接口新增face_enhance参数，支持请求级开关 /health接口新增GFPGAN状态和face_enhance_enabled信息	2026-09-20 17:23:17 +08:00
xiaoxia	8ebccd61d9	fix: 集成真实MuseTalk推理 + TTS style/speed/volume参数全链路修复 ... MuseTalk推理修复（P0）: - 替换_run_inference() stub代码为真实MuseTalk推理逻辑 - 新增音频预处理：任意格式→16kHz mono 16bit WAV - 模型懒加载（VAE+UNet+PE+Whisper），首次推理后复用 - 使用mirror indexing循环帧，消除视频循环边界跳变 - bbox_shift参数可通过请求配置（范围-5~5） - FP16推理支持，节省RTX2060显存 - /health接口增加模型加载状态信息 TTS参数透传修复（P0）: - schemas新增style字段（TTSSynthesizeRequest/TTSPreviewRequest/CreateLipsyncJobRequest/AiAvatarTtsPreviewRequest） - schemas新增volume字段（TTSSynthesizeRequest/CreateLipsyncJobRequest） - cosyvoice_service新增STYLE_INSTRUCTION_MAP和build_style_instruction() - submit_synthesize_task新增style/pitch参数支持 - workflow start_synthesis/resynthesize/submit_segments全链路传递style/volume/pitch - tts.py路由/lipsync.py路由/lipsync_service.py全链路透传新参数 - lipsync_tts.py Celery task支持style/volume参数	2026-09-20 15:52:53 +08:00
xiaoxia	38ffa0b98b	fix(1978): register lipsync_gpu task in worker celery imports ... Worker celery_app.conf.imports was missing app.tasks.lipsync_gpu, so lipsync_gpu_process_async.apply_async() messages were sent to broker but never consumed by any worker, leaving lipsync_jobs stuck at status=processing forever.	2026-09-20 13:44:10 +08:00
xiaoxia	c9876d70e4	fix(1978): fix xiaoxia-gpu-worker.service bad unit file & log path permissions ... - User=%i -> User=ying (%%i is template-only specifier, not valid here) - After=musetalk.service -> After=musetalk-worker.service (match actual service name) - Log path from /tmp/ to $HOME/ to avoid permission issues for ying user	2026-09-20 13:36:36 +08:00
xiaoxia	8413713315	feat(web): TTS配音情感风格选择器 (#2000) ... feat(web): add TTS emotion style selector (#2000)	2026-09-20 12:52:47 +08:00
xiaoxia	9d8c6260e3	perf: async GPU lipsync inference + fix 16x performance regression (#1997) (#1998)	2026-09-20 12:19:35 +08:00
xiaoxia	0764a7820c	Merge pull request 'feat(deploy): GPU节点自动部署配置文件入库（musetalk-worker+gpu-poll+setup/update脚本）' (#1999) from feat/gpu-node-auto-deploy into develop	2026-09-20 11:02:08 +08:00
xiaoxia	db621b4fcb	feat(deploy): #1978 GPU节点自动部署配置文件入库 ... 新增 deploy/gpu_worker/ 下 3 个 systemd 单元 + 3 个运维脚本： - musetalk-worker.service: MuseTalk 推理 API systemd 服务 - gpu-poll.service + gpu-poll.timer: 每30秒轮询Gitea最新commit - scripts/update-gpu-worker.sh: 备份→拉代码→重启→健康检查→失败回滚 - scripts/poll_and_update.sh: SHA比对触发update - scripts/setup-gpu-node.sh: 新节点一键初始化(apt依赖+目录+systemd+sudo免密+首次启动) - README.md 追加第七章「自动部署」说明服务架构/部署步骤/更新机制/日志/注意事项 不改动现有 gpu_worker.py/musetalk_server.py/requirements.txt/xiaoxia-gpu-worker.service/.env.example 不改动CI/Docker/镜像构建。脚本路径写死/home/ying，后续多节点再参数化。	2026-09-20 10:28:24 +08:00
xiaoxia	60cacdf280	style: black formatting	2026-09-20 10:10:54 +08:00
xiaoxia	08de0d9946	perf: async GPU lipsync inference + fix 16x performance regression ... - Move GPU wait_for_result to Celery background task (lipsync_gpu_process_async) POST /lipsync/jobs now returns <1s instead of blocking 200s+ - Rewrite musetalk_server.py: MuseTalk receives full audio directly (v2 architecture) — no pre-looping video before inference Output video length = audio length, mux is fast stream copy - Frontend polls GET /lipsync/jobs/{id} for status updates - refresh_job_status: GPU async path (processing + no mediakit_task_id) skips MediaKit polling; stale jobs (>30min) auto-marked failed - 21 unit tests pass (11 GPU integration + 10 musetalk audio mux) Co-Authored-By: Coze <coze-opensource@bytedance.com>	2026-09-20 09:43:47 +08:00
xiaoxia	b0b81a5d60	fix(1978): MuseTalk 封装强制替换为 TTS 驱动音轨 + 音频长于视频时循环画面 (#1997) ... Co-authored-by: backend-dev <dev@xiaoxiajianji.com> Co-committed-by: backend-dev <dev@xiaoxiajianji.com>	2026-09-20 02:27:44 +08:00
xiaoxia	d959dd874f	feat(1895): 暂停积分系统 ENABLE_CREDIT_SYSTEM=false（保留全部代码/表/接口） (#1996) ... Co-authored-by: backend-dev <dev@xiaoxiajianji.com> Co-committed-by: backend-dev <dev@xiaoxiajianji.com>	2026-09-20 01:39:02 +08:00
xiaoxia	dcd0c56827	feat(web): 暂停积分板块UI展示，保留代码 (ENABLE_CREDIT_SYSTEM=false) (#1995) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-20 01:17:50 +08:00
xiaoxia	585bab9313	fix(ai-avatar): bump lipsync job creation timeout to 120s (#1994) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-20 00:43:08 +08:00
xiaoxia	4a449ae496	fix(1978): GPU lipsync 结果 URL 签 7 天 + 外部 TTS 音频转存自家 OSS (#1993) ... Co-authored-by: backend-dev <dev@xiaoxiajianji.com> Co-committed-by: backend-dev <dev@xiaoxiajianji.com>	2026-09-20 00:07:36 +08:00
xiaoxia	112f0eb277	feat(gpu): #1978 AI数字人口型同步接入MuseTalk GPU Worker（业务侧集成） (#1991)	2026-09-19 22:43:57 +08:00
xiaoxia	2e2d1cd73e	Merge pull request 'fix(gpu): #1970 重写 MuseTalk 服务端 + 客户端超时取消，修复 8 项工程 bug' (#1992) from fix/1970-musetalk-server-rewrite into develop	2026-09-19 20:05:13 +08:00
xiaoxia	32473485d7	fix(gpu): #1970 重写 MuseTalk 服务端 + 客户端超时取消，修复 8 项工程 bug ... 服务端新建 deploy/gpu_worker/musetalk_server.py（替代原 ~/projects/MuseTalk/worker.py）： 1. Flask app.run(threaded=True)：推理阻塞时 /health 仍可达 2. _get_video_fps 兜底：ffprobe 返回 0 或失败时 fallback 到 default_fps(25) 3. _run_ffmpeg 统一封装：subprocess.run(check=True) + timeout，失败/超时抛 RuntimeError 4. inference_lock 并发锁：多请求同时到达时第二请求立即 503 5. 推理超时控制：thread.join(timeout=inference_timeout) 默认 600s，超时返回 504 6. finally 块清理临时目录：成功/失败/超时都删除 task_dir 7. 无人脸检测兜底：_run_inference 中帧提取后校验，无帧直接抛错返回 500 8. 上传大小限制：视频 <=100MB / 音频 <=20MB，超限返回 413 9. 新增 POST /cancel 端点：终止当前推理、清理临时文件、释放锁 10. GET /health 返回 GPU 显存信息（nvidia-smi）+ 当前任务状态 客户端 deploy/gpu_worker/gpu_worker.py 配套： - _call_musetalk 超时后 POST /cancel 终止服务端僵尸推理 - _call_musetalk 返回 (ok, duration, err, retryable) 四元组 - _handle_task 仅 retryable=True 时重试，4xx/短视频等确定性失败直接上报 - 新增 _cancel_musetalk_task 辅助方法 测试：新增 15 个单测覆盖服务端全部修复点；全量 15854 passed / 28 skipped 部署提醒：用户需在 RTX2060 上 wget 新 musetalk_server.py 替换旧 worker.py 并重启服务。	2026-09-19 19:35:23 +08:00
xiaoxia	1591259bb8	fix(gpu): #1970 MuseTalk worker 推理期心跳/超时900/重试收敛/短视频前置失败 (#1990) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-19 15:06:51 +08:00
xiaoxia	a1f25a4426	feat(worker): #1970 AI 标签 backfill 支持 force 重打降级记录 (#1989) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-19 10:49:21 +08:00
xiaoxia	65a77e3fb6	Merge branch 'feat/add-doubao-vision-model-env' into develop ... # Conflicts: # scripts/render_env.sh	2026-09-19 10:32:46 +08:00
xiaoxia	9a57b0d5b8	feat: add DOUBAO_VISION_MODEL env to staging/production deployment ... - Add DOUBAO_VISION_MODEL placeholder to .env.staging and .env.production - Add DOUBAO_VISION_MODEL to render_env.sh SHARED_SECRETS - Add DOUBAO_VISION_MODEL secret ref in ci-pipeline.yml (staging + production) - Uses existing endpoint ep-20260721114705-b568m which supports vision	2026-09-19 10:23:22 +08:00
xiaoxia	4d98e98b57	fix(worker): #1970 注册 AI 标签 Celery 任务（worker.tag_atom_clip unregistered） (#1987) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-19 08:42:07 +08:00
xiaoxia	81e1eb47fb	test(e2e): migrate to asset-libraries + /upload APIs (#1986) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-19 02:11:46 +08:00
xiaoxia	d3e4d6a07d	feat: #1970 hflip 按 atom_clip ai_tags.has_text 放开 + 修复 develop migration 双头 (#1985)	2026-09-19 02:03:59 +08:00
xiaoxia	0d6ce433d0	fix: #1970 删除漏删的重复 migration 081_atom_clip_ai_tags（正确版已编号为 082） (#1984) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-19 01:20:23 +08:00
CI Bot	eb2b009b33	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-18 17:15:42 +00:00
xiaoxia	fbd89b4089	feat: #1970 hflip 按 atom_clip ai_tags.has_text 放开 + 清理重复 081 migration ... - render_adapter 按非 audio 源片段顺序批量查 atom_clip.ai_tags， 仅 has_text 显式 false 标记无文字，其余（未打标签/true/null/查询失败）保守不翻转 - UnifiedRenderService 新增 clip_has_text 注入，None 维持 P1 全保守语义 - 删除残留 081_atom_clip_ai_tags.py（与 GPU PR 的 081 撞号，内容已由 082 承载）， develop alembic 恢复单 head：080→081_add_gpu_lipsync→082_atom_clip_ai_tags - 新增 19 个测试（纯函数混合标记/服务门控/适配器解析/失败回退），全量 15819 passed	2026-09-19 01:06:46 +08:00
xiaoxia	9b50e0696e	test(e2e): 更新冒烟测试适配 #1970 智能剪辑新5步流程 (#1983) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-19 00:58:17 +08:00
xiaoxia	9af73dcd86	fix: #1970 migration 编号冲突修复 081→082 (down_revision 链入 081_add_gpu_lipsync)	2026-09-18 21:27:04 +08:00
xiaoxia	6002f7a5e4	fix(gpu): result接口上报不存在task返回404而非500	2026-09-18 21:25:31 +08:00
xiaoxia	7e88440ca9	feat: #1970 片段级 AI 标签 + 叙事加权匹配 + 冗余核查 (#1981) ... feat: #1970 片段级 AI 标签 + 叙事加权匹配 - atom_clip_tagger.py: MediaKit 抽帧 + 豆包视觉 API 识别 - narrative_match.py: AI 标签加权匹配 (2.0 vs 1.0) - Celery 链式触发 + 批量回填脚本 - migration 081 加 ai_tags 列 - 42 新测试，全量 15796 passed	2026-09-18 21:08:01 +08:00
xiaoxia	fbf8844f25	feat(gpu): #1978 MuseTalk GPU Worker 反向轮询对接（后端API + Worker脚本） (#1979) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-18 19:59:49 +08:00
xiaoxia	34ffe14aae	fix(generate): #1970 Step1 智能降重开关紧贴标题文字 (#1977) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-18 10:39:59 +08:00
xiaoxia	4fa3e4eb92	feat(#1970): 新 API 字段 + 叙事模式 PR3 - assembly_mode/script_id/tts_*/video_ratio (#1976) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-18 07:30:43 +08:00
xiaoxia	a59a6a588a	feat(#1970): 智能降重 PR2 - dedup_enabled 开关 + 6 维片段级微变换 (#1975) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-18 05:44:59 +08:00
xiaoxia	f1621ace9f	feat(#1970): 素材原子化切片 P1 - 数据层/切片逻辑/原子片段级选片 (#1974) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-18 03:57:07 +08:00
xiaoxia	f9daa08b2e	feat(generate): #1970 智能剪辑流程重构 - 选择模式→素材→标题→确认→封面 (#1973)	2026-09-18 00:23:19 +08:00
CI Bot	66409fde6f	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-17 16:00:35 +00:00
xiaoxia	3c016af076	fix(douyin): 本地ASR不可用时正确降级到desc兜底，避免502直接抛出 ... - script_asr_service.py: 捕获ImportError(apps.worker未安装)，抛ASRNotConfiguredError - scripts_ai.py B2回退路径: 504超时直接抛出，502/503 ASR/下载错误走desc兜底 - 修复API镜像未打包worker模块导致有旁白视频在MediaKit失败时直接502的问题 - 更新单元测试覆盖ASR失败→desc兜底场景	2026-09-17 23:48:22 +08:00
xiaoxia	428b9eeb3b	chore(ci): 在CI secrets/env模板中添加TIKHUB_API_KEY和APIZERO_API_KEY ... - Gitea Actions Secrets已添加TIKHUB_API_KEY - render_env.sh SHARED_SECRETS增加两个新key - deploy/configs/.env.staging和.env.production模板增加三层兜底配置段 - ci-pipeline.yml的staging/production deploy步骤env中注入新secret	2026-09-17 22:36:44 +08:00
xiaoxia	43dfd6d425	refactor(douyin): 三层兜底方案最终版（App Feed+TikHub+apizero），移除GoDownloader，清理硬编码Key ... - 移除 P3 GoDownloader/RapidAPI（不再需要） - douyin_resolver.py: P0 App Feed(免费) → P1 TikHub → P2 apizero - API Key 全部从 os.environ 读取，无任何硬编码 - 清理 .env.example 中 RAPIDAPI_KEY 配置项 - scripts_ai.py docstring 更新 - 错误码分阶段（parse/asr/download）保留 - yt-dlp/cookies/ttwid 死代码已在前置提交删除	2026-09-17 21:51:08 +08:00
xiaoxia	73b9f7f97c	chore: retry staging deploy	2026-09-17 20:59:51 +08:00
xiaoxia	ed208ba3f7	chore(douyin): 触发CI重新部署	2026-09-17 20:28:47 +08:00
CI Bot	99d6a47ec5	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-17 11:45:21 +00:00
xiaoxia	9e0959cb85	test(douyin): 更新单元测试适配多源resolver架构，删除yt-dlp/cookies相关mock	2026-09-17 19:30:08 +08:00
CI Bot	8fb75162d4	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-17 10:54:40 +00:00
xiaoxia	01f3e7d4c1	refactor(douyin): 多源轮询解析重构，删除yt-dlp/cookies死代码 ... - 新建 douyin_resolver.py: P0 App Feed API(免费) → P1 TikHub → P2 apizero → P3 GoDownloader - 删除 yt-dlp、cookies、ttwid、HTML直抓等已失效的死代码 - scripts_ai.py 大幅精简：resolver → MediaKit ASR → 本地下载ASR → desc兜底 - 图文视频/无旁白视频统一在resolver层识别并处理 - 错误信息保持分阶段（解析/下载/ASR），不暴露内部细节 - 未配置第三方API Key时P0免费源仍可正常工作 - 新增 TIKHUB_API_KEY/APIZERO_API_KEY/RAPIDAPI_KEY 环境变量 - __debug_diag 接口改为报告各源可用状态	2026-09-17 18:47:52 +08:00