fix: volume=0 clips retained with silence filter to preserve A/V sync

- render_audio.py: stop filtering out volume=0 clips from concat; they produce silence via volume=0 filter, keeping timeline aligned
- unified_render_service.py: probe audio regardless of volume; volume=0 applied as filter, not by dropping stream
- test: update test_silent_clip to verify volume=0 clip retained with volume=0 filter

apps/worker/video_processing/render_audio.py

@@ -135,8 +135,9 @@ def mix_audio(
         audio_clips = layer_map["audio"].clips
 
     # ── 保留源视频原声：过滤掉无音频流的 main clip（图片/无声素材） ──
-    # 每个 clip 按 config.volume 控制音量（0=静音，1=原声），静音 clip 不送入 concat
-    main_clips = [c for c in main_clips if clip_has_audio(ctx, c) and _clip_volume(c) > 0]
+    # 注意：volume=0 的 clip 不能移除——移除会导致后续 clip 音频时间轴前移、音画不同步。
+    # volume=0 通过滤镜链生成静音流，保持时间轴对齐。
+    main_clips = [c for c in main_clips if clip_has_audio(ctx, c)]
 
     # ── 防御：过滤掉无音频流的独立音频轨 ──
     audio_clips = [c for c in audio_clips if clip_has_audio(ctx, c)]

apps/worker/video_processing/unified_render_service.py

@@ -1277,9 +1277,9 @@ class UnifiedRenderService:
         ]
 
         # 音频处理：background 通常是图片无音频，跳过；其他角色先探测是否真有音频流。
-        # volume=0 表示该片段静音，直接丢弃音频（与多片段 mix_audio 行为一致）。
+        # volume=0 不丢弃音频流，而是保留后通过 volume=0 滤镜静音，保持时间轴对齐。
         clip_volume = UnifiedRenderService._clip_volume(clip)
-        if role != "background" and clip_volume > 0:
+        if role != "background":
             try:
                 has_audio = probe_has_audio(clip.local_path)
             except Exception as e:

tests/unit/test_original_audio_retained.py

@@ -123,8 +123,8 @@ class TestOriginalAudioRetained:
         assert "asset_tts1.mp4" in cmd_str
         assert "volume=0.5" in cmd_str
 
-    def test_silent_clip_volume_zero_excluded(self):
-        """volume=0 的素材不送入原声 concat。"""
+    def test_silent_clip_volume_zero_retained_with_silence_filter(self):
+        """volume=0 的素材必须保留在 concat 中（用 volume=0 滤镜静音），不能移除以避免音画不同步。"""
         clips = [_clip("mute", order=0, config={"volume": 0})]
         svc = _service(clips)
         with (
@@ -135,8 +135,11 @@ class TestOriginalAudioRetained:
         ):
             result = mix_audio(_ctx(), _layers(svc, clips), 5.0)
 
-        assert result is None
-        mock_run.assert_not_called()
+        # 有音频流 → 应生成音频文件，且 ffmpeg 命令包含 volume=0.0000 静音滤镜
+        assert result is not None
+        mock_run.assert_called_once()
+        cmd_str = " ".join(mock_run.call_args[0][0])
+        assert "volume=0" in cmd_str
 
     def test_no_audio_stream_returns_none(self):
         clips = [_clip("c1")]