音频文件上传 | AI创作模块兼容音频

video-gen-api/app/api/v1/generation.py

@@ -47,6 +47,11 @@ from app.services.generation_refund_service import mark_generation_record_failed
 from app.services.media_token_usage_snapshot_service import sync_generation_record_media_token_snapshot
 from app.services.credit_record_meta_service import build_generation_record_prompt_meta
 from app.services.video_cover_service import async_create_video_cover_for_local_video
+from app.enums.audio_reference import (
+    AUDIO_ALLOWED_EXTENSIONS,
+    AUDIO_ALLOWED_MIME_TYPES,
+    AUDIO_MAX_FILE_SIZE_MB,
+)
 from app.utils.id_gen import generate_id
 from app.utils.exceptions import InsufficientCreditsError, RecordNotFoundError, InvalidStatusError
 
@@ -832,6 +837,44 @@ async def upload_video(
     return {"url": url, "filename": file.filename or safe_name, "type": "video"}
 
 
+@router.post("/upload-audio")
+async def upload_audio(
+    file: UploadFile = File(...),
+    current_user: User = Depends(get_current_user),
+):
+    """Upload an audio file for AI creation reference."""
+    import os
+    import uuid
+    from app.config import settings
+    from datetime import datetime
+
+    ext = os.path.splitext(file.filename or "")[1].lower().lstrip(".")
+    if ext not in AUDIO_ALLOWED_EXTENSIONS:
+        raise HTTPException(status_code=400, detail="仅支持 mp3、wav 音频文件")
+
+    expected_mime = AUDIO_ALLOWED_MIME_TYPES.get(ext)
+    if not file.content_type or file.content_type != expected_mime:
+        raise HTTPException(status_code=400, detail=f"音频 MIME 类型错误，{ext} 必须为 {expected_mime}")
+
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    safe_name = f"audio_ref_{current_user.id}_{timestamp}_{uuid.uuid4().hex[:8]}.{ext}"
+    date_dir = datetime.now().strftime("%Y/%m/%d")
+    dir_path = os.path.join(settings.UPLOAD_LOCAL_PATH, "audios", date_dir)
+    os.makedirs(dir_path, exist_ok=True)
+    file_path = os.path.join(dir_path, safe_name)
+
+    content = await file.read()
+    max_bytes = AUDIO_MAX_FILE_SIZE_MB * 1024 * 1024
+    if len(content) > max_bytes:
+        raise HTTPException(status_code=400, detail=f"音频大小不能超过{AUDIO_MAX_FILE_SIZE_MB}MB")
+
+    with open(file_path, "wb") as f:
+        f.write(content)
+
+    url = f"/uploads/audios/{date_dir}/{safe_name}"
+    return {"url": url, "filename": file.filename or safe_name, "type": "audio"}
+
+
 @router.post("/delete-file")
 async def delete_upload(
     url: str = Query(..., description="文件URL，如 /uploads/images/2024/01/01/video_img_xxx.png"),

video-gen-api/app/api/v1/video_engines.py

@@ -51,6 +51,7 @@ async def list_active_engines(
             "supported_durations": durations,
             "max_image_count": e.max_image_count,
             "max_video_count": e.max_video_count,
+            "max_audio_count": e.max_audio_count,
             "supports_first_last_frame": e.supports_first_last_frame,
             "supports_universal_reference": e.supports_universal_reference,
         })