MCP ready

app.py

@@ -400,6 +400,32 @@ def random_seed():
 
 @spaces.GPU(duration=180)
 def infer(lq_sequence, task_name, mask, seed, progress=gr.Progress(track_tqdm=True)):
+    """
+    Perform face video restoration on an input video using selected tasks.
+
+    This function supports multiple restoration tasks on low-quality video input, 
+    including Blind Face Restoration (BFR), Colorization, and Inpainting.
+    
+    Depending on user selection, it processes the video and applies the selected 
+    restoration models. It then returns two video outputs:
+    - The face region extracted from the input
+    - The fully restored output video
+    
+    Args:
+        lq_sequence: The input low-quality video file (can be a path or Gradio video object).
+        task_name: A list of selected restoration tasks to perform. Choices include:
+            - "BFR" for Blind Face Restoration
+            - "Colorization" for adding color to grayscale footage
+            - "Inpainting" for filling in missing or corrupted regions using a mask
+        mask: An optional image file that defines regions to be inpainted (only needed if "Inpainting" is selected).
+        seed: An integer seed for random number generation to ensure reproducibility.
+        progress: (Optional) A Gradio progress object for UI progress tracking (used automatically).
+
+    Returns:
+        A tuple of:
+            - face_region_video_path: Path to the video showing extracted face region
+            - output_video_path: Path to the restored output video
+    """
     
     if is_shared_ui:
         lq_sequence, temp_input_folder = process_video(lq_sequence)
@@ -527,4 +553,4 @@ with gr.Blocks(css=css) as demo:
         outputs=[input_seq, task_name, mask_input, seed_input, output_face, output_res]
     )
 
-demo.queue().launch(show_api=False, show_error=True, ssr_mode=False)
+demo.queue().launch(show_api=False, show_error=True, ssr_mode=False, mcp_server=True)