fix(#549): 预设配音无声 - 字幕对齐TTS配音模式 + ASR缓存 (#628)

Co-authored-by: xiaoxia <dev@xiaoxiajianji.com>
Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>

apps/worker/video_processing/unified_render_service.py

@@ -188,6 +188,8 @@ class UnifiedRenderService:
         self.bgm_path = bgm_path
         self._transition_engine = TransitionEngine(default_duration=transition_duration)
         self._speed_engine = SpeedEngine()
+        self._asr_timeline_cache: Any = None  # ASR 字幕结果缓存，避免重复调用
+        self._asr_timeline_cached = False
 
     def render(self) -> RenderResult:
         """执行渲染，返回 RenderResult.
@@ -589,7 +591,13 @@ class UnifiedRenderService:
 
         MVP 版本：使用第一个有音频的素材做ASR，然后按比例映射到整个视频时长。
         后续优化：支持多片段拼接后的完整音频ASR。
+
+        带缓存：同一 plan 只做一次 ASR，TTS 配音和字幕共用结果。
         """
+        # 检查缓存
+        if self._asr_timeline_cached:
+            return self._asr_timeline_cache
+
         from packages.domain.subtitle import SubtitleTimeline
 
         # 找第一个有本地路径的素材
@@ -602,7 +610,10 @@ class UnifiedRenderService:
 
         if first_asset_path is None:
             logger.warning("ASR字幕生成失败：找不到可用素材音频")
-            return SubtitleTimeline(segments=[], total_duration=video_duration)
+            result = SubtitleTimeline(segments=[], total_duration=video_duration)
+            self._asr_timeline_cache = result
+            self._asr_timeline_cached = True
+            return result
 
         # 提取素材音频为 wav（16kHz单声道，ASR友好格式）
         audio_path = self.work_dir / f"asr_audio_{self.plan.id}.wav"
@@ -637,6 +648,9 @@ class UnifiedRenderService:
         except Exception:
             pass
 
+        # 存入缓存
+        self._asr_timeline_cache = timeline
+        self._asr_timeline_cached = True
         return timeline
 
     def _extract_audio(self, video_path: Path, output_path: Path) -> None:
@@ -669,11 +683,54 @@ class UnifiedRenderService:
     ) -> bool:
         """根据 plan.config 生成 TTS 配音，加到 audio 图层.
 
+        支持三种触发方式：
+        1. config.tts.enabled = true → 标准 TTS 配置
+        2. 顶层 voice_id + custom_text → 桥接模式（自定义文案配音）
+        3. 顶层 voice_id + subtitle.auto_generated=true → ASR 字幕对齐配音（预设配音）
+
         Returns:
             是否成功添加了配音音轨
         """
         config = self.plan.config or {}
         tts_cfg = config.get("tts", {}) or {}
+        subtitle_cfg = config.get("subtitle", {}) or {}
+        use_subtitle_align = False  # 是否使用字幕对齐模式
+
+        # 兼容前端顶层字段：voice_id / custom_text / voice_clone_profile_id
+        if not tts_cfg.get("enabled"):
+            top_voice_id = config.get("voice_id", "") or ""
+            top_text = config.get("custom_text", "") or ""
+
+            # 方式A：voice_id + custom_text → 整段配音
+            if top_voice_id and top_text:
+                tts_cfg = {
+                    "enabled": True,
+                    "voice_id": top_voice_id,
+                    "text": top_text,
+                    "align_mode": "full",
+                    "overlap_mode": "replace",
+                }
+                logger.info(
+                    "[unified-render] 检测到顶层 voice_id+custom_text，桥接到 tts 配置（整段）: plan_id=%s voice_id=%s text_len=%d",
+                    self.plan.id,
+                    top_voice_id,
+                    len(top_text),
+                )
+            # 方式B：voice_id + 自动字幕 → 字幕对齐配音（预设配音模式）
+            elif top_voice_id and subtitle_cfg.get("auto_generated", False) and self.asr_service is not None:
+                tts_cfg = {
+                    "enabled": True,
+                    "voice_id": top_voice_id,
+                    "text": "",
+                    "align_mode": "subtitle",
+                    "overlap_mode": "replace",
+                }
+                use_subtitle_align = True
+                logger.info(
+                    "[unified-render] 检测到预设配音+自动字幕，使用字幕对齐模式: plan_id=%s voice_id=%s",
+                    self.plan.id,
+                    top_voice_id,
+                )
 
         # 兼容前端顶层字段：voice_id / custom_text / voice_clone_profile_id
         # 前端一键生成页面传 config.voice_id + config.custom_text，
@@ -706,8 +763,34 @@ class UnifiedRenderService:
             tts_service = get_tts_service()
             tts_engine = TtsEngine(tts_service, self.work_dir / "tts")
 
-            # 整段配音模式
-            result = tts_engine.generate_full_voiceover(tts_config, total_duration=video_duration)
+            # 根据对齐模式选择生成方式
+            if use_subtitle_align or tts_config.align_mode == "subtitle":
+                # 字幕对齐模式：先做 ASR，再按字幕生成配音
+                if not self._asr_timeline_cached:
+                    self._generate_asr_subtitles(video_duration, subtitle_cfg)
+                timeline = self._asr_timeline_cache
+                if timeline is None or not timeline.segments:
+                    logger.warning("TTS 字幕对齐配音：ASR 无识别结果，跳过配音")
+                    return False
+
+                # 转换为 TtsEngine 需要的字幕格式
+                subtitles = [
+                    {
+                        "text": seg.text,
+                        "start_time": seg.start,
+                        "end_time": seg.end,
+                    }
+                    for seg in timeline.segments
+                    if getattr(seg, "text", "").strip()
+                ]
+                if not subtitles:
+                    logger.warning("TTS 字幕对齐配音：字幕文本为空，跳过配音")
+                    return False
+
+                result = tts_engine.generate_subtitle_voiceover(tts_config, subtitles)
+            else:
+                # 整段配音模式
+                result = tts_engine.generate_full_voiceover(tts_config, total_duration=video_duration)
 
             if not result.success or not result.segments:
                 logger.warning("TTS 配音生成失败，跳过: %s", result.error_message)