音频文件上传 | AI创作模块兼容音频

video-gen-api/app/services/generation_ai_service.py

@@ -15,6 +15,13 @@ from app.models.project import Project
 from app.models.image_engine import ImageEngine
 from app.models.user import User
 from app.models.video_engine import VideoEngine
+from app.enums.audio_reference import (
+    AUDIO_ALLOWED_EXTENSIONS,
+    AUDIO_MAX_COUNT_LIMIT,
+    AUDIO_MAX_DURATION_SECONDS,
+    AUDIO_MAX_TOTAL_DURATION_SECONDS,
+    AUDIO_MIN_DURATION_SECONDS,
+)
 from app.enums.generation_history import (
     GenerationHistorySourceEnum,
     get_generation_history_source_label,
@@ -157,6 +164,7 @@ def _build_video_snapshot(engine: VideoEngine, ratio: str, resolution: str, dura
         "supported_resolutions": _parse_list(engine.supported_resolutions, []),
         "supported_durations": _parse_list(engine.supported_durations, []),
         "max_duration": engine.max_duration,
+        "max_audio_count": engine.max_audio_count,
         "selected_ratio": ratio,
         "selected_resolution": resolution,
         "selected_duration": duration,
@@ -203,6 +211,7 @@ async def list_generation_ai_engine_options(db: AsyncSession) -> GenerationAIEng
             priority=engine.priority or 0,
             max_image_count=engine.max_image_count,
             max_video_count=engine.max_video_count,
+            max_audio_count=engine.max_audio_count,
             supports_first_last_frame=engine.supports_first_last_frame,
             supports_universal_reference=engine.supports_universal_reference,
         )
@@ -244,6 +253,8 @@ async def create_async_generation_task(db: AsyncSession, current_user: User, req
     await assert_user_resource_capacity_available(db, current_user.id)
 
     if gen_type == "image":
+        if any((r.get("type") or "").lower() == "audio" for r in refs):
+            raise HTTPException(status_code=400, detail="图片生成不支持音频参考素材")
         engine = await _get_image_engine(db, req.engine_id)
         sizes = _image_supported_sizes(engine)
         size = req.image_size or engine.default_size or IMAGE_DEFAULT_SIZE
@@ -306,7 +317,7 @@ async def create_async_generation_task(db: AsyncSession, current_user: User, req
 
         input_video_duration = 0.0
         if refs:
-            video_refs = [r for r in refs if r.get("type") == "video"]
+            video_refs = [r for r in refs if (r.get("type") or "").lower() == "video"]
             for ref in video_refs:
                 ref_duration = float(ref.get("duration") or 0)
                 if ref_duration < 2:
@@ -315,6 +326,42 @@ async def create_async_generation_task(db: AsyncSession, current_user: User, req
             if input_video_duration > 15:
                 raise HTTPException(status_code=400, detail=f"所有视频素材总时长不能超过 15 秒，当前 {input_video_duration:.1f} 秒")
 
+            audio_refs = [r for r in refs if (r.get("type") or "").lower() == "audio"]
+            if audio_refs:
+                max_audio_count = int(engine.max_audio_count or 0)
+                if max_audio_count <= 0:
+                    raise HTTPException(status_code=400, detail="当前视频引擎不支持音频参考素材")
+                if max_audio_count > AUDIO_MAX_COUNT_LIMIT:
+                    max_audio_count = AUDIO_MAX_COUNT_LIMIT
+                if len(audio_refs) > max_audio_count:
+                    raise HTTPException(
+                        status_code=400,
+                        detail=f"参考音频最多可传 {max_audio_count} 段，当前 {len(audio_refs)} 段",
+                    )
+
+                input_audio_duration = 0.0
+                for ref in audio_refs:
+                    raw_duration = ref.get("duration")
+                    if raw_duration is None:
+                        raw_duration = 0.0
+                    try:
+                        ref_duration = float(raw_duration)
+                    except (TypeError, ValueError):
+                        ref_duration = 0.0
+
+                    if ref_duration < AUDIO_MIN_DURATION_SECONDS or ref_duration > AUDIO_MAX_DURATION_SECONDS:
+                        raise HTTPException(
+                            status_code=400,
+                            detail=f"单段参考音频时长必须在 {AUDIO_MIN_DURATION_SECONDS}-{AUDIO_MAX_DURATION_SECONDS} 秒之间",
+                        )
+                    input_audio_duration += ref_duration
+
+                if input_audio_duration > AUDIO_MAX_TOTAL_DURATION_SECONDS:
+                    raise HTTPException(
+                        status_code=400,
+                        detail=f"所有参考音频总时长不能超过 {AUDIO_MAX_TOTAL_DURATION_SECONDS} 秒，当前 {input_audio_duration:.1f} 秒",
+                    )
+
         media_billing = await charge_generation_media_by_params(
             db,
             user_id=current_user.id,

video-gen-api/app/services/video_gen.py

@@ -135,6 +135,10 @@ async def submit_video_task(
                     resolved = _resolve_url(ref_url)
                     role = ref_role if ref_role else "reference_video"
                     content.append({"type": "video_url", "video_url": {"url": resolved}, "role": role})
+                elif ref_type == "audio" and ref_url:
+                    resolved = _resolve_url(ref_url)
+                    role = ref_role if ref_role else "reference_audio"
+                    content.append({"type": "audio_url", "audio_url": {"url": resolved}, "role": role})
         except (json.JSONDecodeError, TypeError):
             pass