Add target_fps parameter to video task requests and update inference worker (#528)

lightx2v/server/api/tasks/video.py

@@ -52,6 +52,7 @@ async def create_video_task_form(
     seed: int = Form(default=42),
     audio_file: UploadFile = File(None),
     video_duration: int = Form(default=5),
+    target_fps: int = Form(default=16),
 ):
     services = get_services()
     assert services.file_service is not None, "File service is not initialized"
@@ -89,6 +90,7 @@ async def create_video_task_form(
         seed=seed,
         audio_path=audio_path,
         video_duration=video_duration,
+        target_fps=target_fps,
     )
 
     try:

lightx2v/server/schema.py

@@ -40,6 +40,7 @@ class VideoTaskRequest(BaseTaskRequest):
     audio_path: str = Field("", description="Input audio path (Wan-Audio)")
     video_duration: int = Field(5, description="Video duration (Wan-Audio)")
     talk_objects: Optional[list[TalkObject]] = Field(None, description="Talk objects (Wan-Audio)")
+    target_fps: Optional[int] = Field(16, description="Target FPS for video frame interpolation (overrides config)")
 
 
 class ImageTaskRequest(BaseTaskRequest):
@@ -53,6 +54,7 @@ class TaskRequest(BaseTaskRequest):
     video_duration: int = Field(5, description="Video duration (Wan-Audio)")
     talk_objects: Optional[list[TalkObject]] = Field(None, description="Talk objects (Wan-Audio)")
     aspect_ratio: str = Field("16:9", description="Output aspect ratio (T2I only)")
+    target_fps: Optional[int] = Field(16, description="Target FPS for video frame interpolation (overrides config)")
 
 
 class TaskStatusMessage(BaseModel):

lightx2v/server/services/inference/worker.py

@@ -55,6 +55,11 @@ class TorchrunInferenceWorker:
             task_data["return_result_tensor"] = False
             task_data["negative_prompt"] = task_data.get("negative_prompt", "")
 
+            if task_data.get("target_fps") is not None and "video_frame_interpolation" in self.runner.config:
+                task_data["video_frame_interpolation"] = dict(self.runner.config["video_frame_interpolation"])
+                task_data["video_frame_interpolation"]["target_fps"] = task_data["target_fps"]
+                del task_data["target_fps"]
+
             task_data = EasyDict(task_data)
             input_info = set_input_info(task_data)