Update app.py

app.py

@@ -201,26 +201,12 @@ def generate_video(gallery_images, mode, prompt, height, width,
         reference_images = None
     elif mode == "Ref2V": 
         frames, mask = prepare_video_and_mask_Ref2V(height=target_h, width=target_w, num_frames=num_frames)
-        # Resize reference images to match target dimensions
-        reference_images = [img.resize((target_w, target_h)) for img in gallery_images]
+        reference_images = gallery_images
     else:  # mode == "Random2V"
-        # Calculate appropriate frame indices based on number of images and frames
-        num_images = len(gallery_images)
-        if num_images == 1:
-            frame_indices = [num_frames // 2]  # Place single image in the middle
-        elif num_images == 2:
-            frame_indices = [0, num_frames - 1]  # Place at start and end
-        else:
-            # Distribute images evenly across the video
-            # Ensure we don't exceed available frames
-            max_images = min(num_images, num_frames)
-            step = max(1, num_frames // max_images)
-            frame_indices = [min(i * step, num_frames - 1) for i in range(max_images)]
-            gallery_images = gallery_images[:max_images]  # Limit images to what we can use
         
         frames, mask = prepare_video_and_mask_Random2V(
             images=gallery_images, 
-            frame_indices=frame_indices, 
+            frame_indices=[0,20,40], # todo - generalize
             height=target_h, 
             width=target_w, 
             num_frames=num_frames
@@ -298,7 +284,7 @@ with gr.Blocks() as demo:
             
         with gr.Column():
             video_output = gr.Video(label="Generated Video", autoplay=True, interactive=False)
-            with gr.Accordion("Mode Information", open=True):
+            with gr.Accordion("Mode Information", open=False):
                 gr.Markdown("""
                 **Processing Modes:**
                 - **Ref2V**: Uses uploaded images as style references for video generation. All frames are generated based on the reference images.
@@ -331,4 +317,4 @@ with gr.Blocks() as demo:
     generate_button.click(fn=generate_video, inputs=ui_inputs, outputs=[video_output, seed_input])
 
 if __name__ == "__main__":
-    demo.queue().launch()
+    demo.queue().launch(mcp_server=True)