apps/worker/video_processing/unified_render_service.py

@@ -1494,9 +1494,17 @@ class UnifiedRenderService:
             raise ValueError("没有可渲染的图层")
 
         # 收集所有 clips（按图层顺序，同层按 order）
+        # 排除纯音频 clips — 它们由 mix_audio() 独立处理，不应出现在视频 filter_complex 中
+        # 例如：voice.mp3 没有视频流，如果加入 all_clips 会生成 [N:v] 引用导致 FFmpeg 报错
         all_clips: list[ResolvedClip] = []
         for layer in layers:
-            all_clips.extend(layer.clips)
+            for clip in layer.clips:
+                if clip.clip_type == "audio":
+                    continue
+                all_clips.append(clip)
+
+        if not all_clips:
+            raise ValueError("没有可渲染的视频片段（所有片段均为纯音频）")
 
         # 构建输入参数
         input_args: list[str] = []
@@ -1580,10 +1588,14 @@ class UnifiedRenderService:
             filter_parts.append(filter_str)
             preprocessed_labels.append(label)
 
-        # Step 2: 同层 clips 用 xfade 串联
+        # Step 2: 同层 clips 用 xfade 串联（跳过纯音频层，由 mix_audio() 独立处理）
         layer_output_labels: dict[str, str] = {}
         for layer in layers:
-            layer_clip_indices = [all_clips.index(c) for c in layer.clips]
+            # 音频层不参与视频 filter_complex，跳过
+            video_clips_in_layer = [c for c in layer.clips if c.clip_type != "audio"]
+            if not video_clips_in_layer:
+                continue
+            layer_clip_indices = [all_clips.index(c) for c in video_clips_in_layer]
             layer_labels = [preprocessed_labels[i] for i in layer_clip_indices]
             # 使用调速后的实际时长，与 Step 1 的调速处理保持一致
             layer_durations = [UnifiedRenderService._clip_adjusted_duration(all_clips[i]) for i in layer_clip_indices]

tests/unit/test_audio_only_clips_in_filter_complex.py

@@ -0,0 +1,262 @@
+"""测试 _build_filter_complex 正确排除纯音频 clips.
+
+Bug: voice.mp3（纯音频文件）被错误地加入视频 filter_complex，
+导致 FFmpeg 尝试访问 [N:v] 视频流时报错 "Stream specifier ':v' matches no streams".
+
+修复：_build_filter_complex 在收集 clips 时跳过 clip_type="audio" 的 clips，
+因为音频 clips 由 mix_audio() 独立处理。
+"""
+
+from pathlib import Path
+from unittest.mock import MagicMock
+
+import pytest
+
+
+def _make_render_service(tmp_path):
+    """创建一个最小化的 UnifiedRenderService 实例."""
+    from video_processing.unified_render_service import UnifiedRenderService
+
+    plan = MagicMock()
+    plan.id = "test_plan"
+    plan.config = {}
+    plan.strategy_id = "test_strategy"
+
+    clips = []
+    asset_path_map = {}
+
+    service = UnifiedRenderService(
+        plan=plan,
+        clips=clips,
+        asset_path_map=asset_path_map,
+        work_dir=tmp_path,
+        output_width=480,
+        output_height=854,
+        output_fps=30,
+    )
+    return service
+
+
+def _make_video_clip(clip_id: str, local_path: Path, duration: float = 5.0):
+    """创建一个视频 clip."""
+    from video_processing.unified_render_service import ResolvedClip
+
+    return ResolvedClip(
+        clip_id=clip_id,
+        asset_id=f"asset_{clip_id}",
+        local_path=local_path,
+        clip_type="video",
+        order=0,
+        start_time=0.0,
+        duration=duration,
+        config={},
+        actual_duration=duration,
+    )
+
+
+def _make_audio_clip(clip_id: str, local_path: Path, duration: float = 5.0):
+    """创建一个纯音频 clip."""
+    from video_processing.unified_render_service import ResolvedClip
+
+    return ResolvedClip(
+        clip_id=clip_id,
+        asset_id=f"asset_{clip_id}",
+        local_path=local_path,
+        clip_type="audio",
+        order=0,
+        start_time=0.0,
+        duration=duration,
+        config={"volume": 1.0},
+        actual_duration=duration,
+    )
+
+
+class TestBuildFilterComplexExcludesAudioClips:
+    """_build_filter_complex 应该排除 clip_type='audio' 的 clips."""
+
+    def test_audio_clip_not_in_filter_complex(self, tmp_path):
+        """纯音频 clip 不应出现在 filter_complex 中."""
+        render_service = _make_render_service(tmp_path)
+        from video_processing.unified_render_service import RenderLayer
+
+        # 准备视频和音频文件
+        video_path = tmp_path / "video.mp4"
+        video_path.write_bytes(b"\x00")
+        audio_path = tmp_path / "voice.mp3"
+        audio_path.write_bytes(b"\x00")
+
+        video_clip = _make_video_clip("clip_0", video_path, duration=5.0)
+        audio_clip = _make_audio_clip("voice_library_main", audio_path, duration=5.0)
+
+        video_layer = RenderLayer(role="main", z_index=1)
+        video_layer.clips.append(video_clip)
+
+        audio_layer = RenderLayer(role="audio", z_index=2)
+        audio_layer.clips.append(audio_clip)
+
+        layers = [video_layer, audio_layer]
+
+        # 执行
+        filter_complex, input_args = render_service._build_filter_complex(layers)
+
+        # 验证：filter_complex 只包含视频 clip 的处理（[0:v]），不包含音频 clip（[1:v]）
+        assert "[0:v]" in filter_complex
+        assert "[1:v]" not in filter_complex  # voice.mp3 不应该有视频滤镜
+
+        # 验证：input_args 只包含视频文件，不包含音频文件
+        assert str(video_path) in " ".join(input_args)
+        assert str(audio_path) not in " ".join(input_args)
+
+    def test_multiple_video_clips_with_audio(self, tmp_path):
+        """多个视频 clips + 音频 clip 时，filter_complex 只处理视频."""
+        render_service = _make_render_service(tmp_path)
+        from video_processing.unified_render_service import RenderLayer
+
+        # 准备文件
+        video_paths = [tmp_path / f"video_{i}.mp4" for i in range(3)]
+        for p in video_paths:
+            p.write_bytes(b"\x00")
+        audio_path = tmp_path / "voice.mp3"
+        audio_path.write_bytes(b"\x00")
+
+        video_layer = RenderLayer(role="main", z_index=1)
+        for i, vp in enumerate(video_paths):
+            clip = _make_video_clip(f"clip_{i}", vp, duration=3.0)
+            clip.order = i
+            video_layer.clips.append(clip)
+
+        audio_layer = RenderLayer(role="audio", z_index=2)
+        audio_clip = _make_audio_clip("voice_main", audio_path, duration=9.0)
+        audio_layer.clips.append(audio_clip)
+
+        layers = [video_layer, audio_layer]
+
+        # 执行
+        filter_complex, input_args = render_service._build_filter_complex(layers)
+
+        # 验证：只有 3 个视频输入
+        assert "[0:v]" in filter_complex
+        assert "[1:v]" in filter_complex
+        assert "[2:v]" in filter_complex
+        assert "[3:v]" not in filter_complex  # 音频不应该出现
+
+        # 验证：input_args 只有 3 个 -i
+        input_files = [arg for arg in input_args if not arg.startswith("-")]
+        assert len(input_files) == 3
+        assert str(audio_path) not in input_files
+
+    def test_only_audio_clips_raises_error(self, tmp_path):
+        """只有音频 clips 时应该抛出 ValueError."""
+        render_service = _make_render_service(tmp_path)
+        from video_processing.unified_render_service import RenderLayer
+
+        audio_path = tmp_path / "voice.mp3"
+        audio_path.write_bytes(b"\x00")
+
+        audio_layer = RenderLayer(role="audio", z_index=2)
+        audio_clip = _make_audio_clip("voice_main", audio_path, duration=5.0)
+        audio_layer.clips.append(audio_clip)
+
+        layers = [audio_layer]
+
+        with pytest.raises(ValueError, match="没有可渲染的视频片段"):
+            render_service._build_filter_complex(layers)
+
+    def test_tts_audio_clip_excluded(self, tmp_path):
+        """TTS 配音 clip（clip_type='audio', config.tts=True）也不应出现在 filter_complex."""
+        render_service = _make_render_service(tmp_path)
+        from video_processing.unified_render_service import RenderLayer, ResolvedClip
+
+        video_path = tmp_path / "video.mp4"
+        video_path.write_bytes(b"\x00")
+        tts_audio_path = tmp_path / "tts_segment.wav"
+        tts_audio_path.write_bytes(b"\x00")
+
+        video_clip = _make_video_clip("clip_0", video_path, duration=10.0)
+        tts_clip = ResolvedClip(
+            clip_id="tts_0.000",
+            asset_id="tts_voiceover",
+            local_path=tts_audio_path,
+            clip_type="audio",
+            order=0,
+            start_time=0.0,
+            duration=3.0,
+            config={"volume": 1.0, "tts": True},
+            actual_duration=3.0,
+        )
+
+        video_layer = RenderLayer(role="main", z_index=1)
+        video_layer.clips.append(video_clip)
+
+        audio_layer = RenderLayer(role="audio", z_index=2)
+        audio_layer.clips.append(tts_clip)
+
+        layers = [video_layer, audio_layer]
+
+        filter_complex, input_args = render_service._build_filter_complex(layers)
+
+        # TTS 音频不应出现在 filter_complex
+        assert "[1:v]" not in filter_complex
+        input_files = [arg for arg in input_args if not arg.startswith("-")]
+        assert str(tts_audio_path) not in input_files
+
+    def test_video_clip_with_audio_config_not_excluded(self, tmp_path):
+        """clip_type='video' 的 clip 不应被排除（即使它有音频流）."""
+        render_service = _make_render_service(tmp_path)
+        from video_processing.unified_render_service import RenderLayer
+
+        video_path = tmp_path / "video.mp4"
+        video_path.write_bytes(b"\x00")
+
+        video_clip = _make_video_clip("clip_0", video_path, duration=5.0)
+        video_layer = RenderLayer(role="main", z_index=1)
+        video_layer.clips.append(video_clip)
+
+        layers = [video_layer]
+
+        filter_complex, input_args = render_service._build_filter_complex(layers)
+
+        # 视频 clip 应该被处理
+        assert "[0:v]" in filter_complex
+        input_files = [arg for arg in input_args if not arg.startswith("-")]
+        assert str(video_path) in input_files
+
+    def test_voice_library_clip_with_voice_library_flag(self, tmp_path):
+        """voice_library=True 的 clip（来自 _maybe_add_voice_library_layer）应被排除."""
+        render_service = _make_render_service(tmp_path)
+        from video_processing.unified_render_service import RenderLayer, ResolvedClip
+
+        video_path = tmp_path / "video.mp4"
+        video_path.write_bytes(b"\x00")
+        voice_path = tmp_path / "voice.mp3"
+        voice_path.write_bytes(b"\x00")
+
+        video_clip = _make_video_clip("clip_0", video_path, duration=14.0)
+        # 模拟 _maybe_add_voice_library_layer 创建的 clip
+        voice_clip = ResolvedClip(
+            clip_id="voice_library_main",
+            asset_id="voice_library",
+            local_path=voice_path,
+            clip_type="audio",
+            order=0,
+            start_time=0.0,
+            duration=14.0,
+            config={"volume": 1.0, "voice_library": True},
+            actual_duration=14.0,
+        )
+
+        video_layer = RenderLayer(role="main", z_index=1)
+        video_layer.clips.append(video_clip)
+
+        audio_layer = RenderLayer(role="audio", z_index=2)
+        audio_layer.clips.append(voice_clip)
+
+        layers = [video_layer, audio_layer]
+
+        filter_complex, input_args = render_service._build_filter_complex(layers)
+
+        # 只有视频 clip 被处理
+        assert "[0:v]" in filter_complex
+        assert "[1:v]" not in filter_complex
+        # voice.mp3 不在输入中
+        assert str(voice_path) not in " ".join(input_args)