Update app.py

.gitattributes

@@ -36,5 +36,3 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 examples/man_walking.mp4 filter=lfs diff=lfs merge=lfs -text
 examples/leopard.mp4 filter=lfs diff=lfs merge=lfs -text
 examples/woman.mp4 filter=lfs diff=lfs merge=lfs -text
-examples/neon.mp4 filter=lfs diff=lfs merge=lfs -text
-examples/painter.mp4 filter=lfs diff=lfs merge=lfs -text

app.py

@@ -13,53 +13,58 @@ vae = AutoencoderKLWan.from_pretrained(model_id, subfolder="vae", torch_dtype=to
 pipe = LucyEditPipeline.from_pretrained(model_id, vae=vae, torch_dtype=torch.bfloat16)
 pipe.to("cuda")
 
-def calculate_resolution(input_width, input_height, min_dimension=480, max_dimension=832, compatible_round=32):
-    """Calculate optimal resolution preserving aspect ratio within bounds"""
-    # Ensure dimensions are multiples of the compatible rounding
-    def round_to(x, compatible_round):
-        return max(min_dimension, min(max_dimension, int(round(x / compatible_round) * compatible_round)))
-
+def calculate_resolution(input_width, input_height, min_dimension=480, max_dimension=832):
+    """Calculate optimal resolution preserving aspect ratio"""
+    # Ensure dimensions are multiples of 16
+    def round_to_16(x):
+        return int(round(x / 16.0) * 16)
+    
     # Get aspect ratio
     aspect_ratio = input_width / input_height
-
-    # Square videos (aspect ratio close to 1:1)
-    if 0.98 <= aspect_ratio <= 1.02:
+    
+    # Square videos
+    if 0.95 <= aspect_ratio <= 1.05:
         return 640, 640
-
+    
     # Landscape videos (width > height)
     elif aspect_ratio > 1:
-        # Try to use max width
+        # Start with max width
         new_width = max_dimension
-        new_height = new_width / aspect_ratio
-
-        # If height would be too small, use min height
+        new_height = int(new_width / aspect_ratio)
+        
+        # If height is too small, use min height instead
         if new_height < min_dimension:
             new_height = min_dimension
-            new_width = new_height * aspect_ratio
-            # If width exceeds max, clamp it
+            new_width = int(new_height * aspect_ratio)
+            
+            # Clamp width if needed
             if new_width > max_dimension:
                 new_width = max_dimension
-
-        return round_to(new_width, compatible_round), round_to(new_height, compatible_round)
-
-    # Portrait videos (height > width)
+                new_height = int(new_width / aspect_ratio)
+    
+    # Portrait videos (height > width)  
     else:
-        # Try to use max height
+        # Start with max height
         new_height = max_dimension
-        new_width = new_height * aspect_ratio
-
-        # If width would be too small, use min width
+        new_width = int(new_height * aspect_ratio)
+        
+        # If width is too small, use min width instead
         if new_width < min_dimension:
             new_width = min_dimension
-            new_height = new_width / aspect_ratio
-            # If height exceeds max, clamp it
+            new_height = int(new_width / aspect_ratio)
+            
+            # Clamp height if needed
             if new_height > max_dimension:
                 new_height = max_dimension
-
-        return round_to(new_width, compatible_round), round_to(new_height, compatible_round)
-
-
-@spaces.GPU(duration=120)
+                new_width = int(new_height / aspect_ratio)
+    
+    # Round to multiples of 16 and ensure within bounds
+    final_width = round_to_16(max(min_dimension, min(max_dimension, new_width)))
+    final_height = round_to_16(max(min_dimension, min(max_dimension, new_height)))
+    
+    return final_width, final_height
+
+@spaces.GPU(duration=90)
 def process_video(
     video_path,
     prompt,
@@ -73,21 +78,24 @@ def process_video(
 ):
     # Load and preprocess video
     progress(0.2, desc="Loading video...")
-
+    
     # Get video dimensions
     temp_video = load_video(video_path)
     print(len(temp_video))
     if temp_video and len(temp_video) > 0:
         original_width, original_height = temp_video[0].size
-
+        
         # Calculate dimensions
         if auto_resize:
             width, height = calculate_resolution(original_width, original_height)
+            gr.Info(f"Auto-resized from {original_width}x{original_height} to {width}x{height} (preserving aspect ratio)")
         else:
             width, height = manual_width, manual_height
+            if abs((original_width/original_height) - (width/height)) > 0.1:
+                gr.Warning(f"Output aspect ratio ({width}x{height}) differs significantly from input ({original_width}x{original_height}). Video may appear stretched.")
     else:
         raise gr.Error("Could not load video or video is empty")
-
+    
     # Convert video function
     def convert_video(video: List[Image.Image]) -> List[Image.Image]:
         # Ensure we don't exceed the video length
@@ -96,14 +104,15 @@ def process_video(
         # Resize frames
         video_frames = [frame.resize((width, height)) for frame in video_frames]
         return video_frames
-
+    
     # Load video from file path
     video = load_video(video_path, convert_method=convert_video)
-
+    
     # Ensure we have the right number of frames
     if len(video) < num_frames:
+        gr.Warning(f"Video has only {len(video)} frames, using all available frames.")
         num_frames = len(video)
-
+    
     # Generate edited video
     progress(0.5, desc="Generating edited video...")
     output = pipe(
@@ -115,14 +124,14 @@ def process_video(
         num_frames=num_frames,
         guidance_scale=guidance_scale,
     ).frames[0]
-
+    
     # Export to temporary file
     progress(0.9, desc="Exporting video...")
     with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as tmp_file:
         output_path = tmp_file.name
-
+    
     export_to_video(output, output_path, fps=24)
-
+    
     progress(1.0, desc="Complete!")
     return output_path
 
@@ -141,18 +150,18 @@ with gr.Blocks(title="Lucy Edit - Video Editing with Text", css=css) as demo:
   &nbsp;|&nbsp; 📑 <a href="#">arXiv (Coming soon)</a>
   &nbsp;|&nbsp; 💬 <a href="https://discord.gg/decart">Discord</a>
 </p>""")
-
+    
     with gr.Row():
         with gr.Column(scale=1):
             # Input controls
             video_input = gr.Video(label="Input Video")
-
+            
             prompt = gr.Textbox(
                 label="Edit Prompt",
                 placeholder="Describe what you want to change in the video...",
                 lines=3
             )
-
+            
             with gr.Accordion("Advanced Settings", open=False):
                 negative_prompt = gr.Textbox(
                     label="Negative Prompt (optional)",
@@ -164,7 +173,7 @@ with gr.Blocks(title="Lucy Edit - Video Editing with Text", css=css) as demo:
                     value=True,
                     info="Automatically calculate dimensions based on input video"
                 )
-
+                
                 num_frames = gr.Slider(
                     label="Number of Frames",
                     minimum=1,
@@ -173,7 +182,7 @@ with gr.Blocks(title="Lucy Edit - Video Editing with Text", css=css) as demo:
                     step=1,
                     info="More frames = longer processing time"
                 )
-
+                
                 with gr.Row():
                     manual_height = gr.Slider(
                         label="Height (when auto-resize is off)",
@@ -189,7 +198,7 @@ with gr.Blocks(title="Lucy Edit - Video Editing with Text", css=css) as demo:
                         value=832,
                         step=32
                     )
-
+                
                 guidance_scale = gr.Slider(
                     label="Guidance Scale",
                     minimum=1.0,
@@ -198,23 +207,24 @@ with gr.Blocks(title="Lucy Edit - Video Editing with Text", css=css) as demo:
                     step=0.5,
                     info="Higher values follow the prompt more strictly"
                 )
-
+            
             generate_btn = gr.Button("Edit Video", variant="primary")
-
+            
         with gr.Column(scale=1):
-            video_output = gr.Video(label="Edited Video", autoplay=True)
+            video_output = gr.Video(label="Edited Video")
 
     gr.Examples(
-        examples=[
-            ["examples/neon.mp4", "Add a colorful scarlet macaw parrot perched on the man's left shoulder, bright red and blue wing feathers with yellow accents, curved black beak, intelligent dark eyes, talons gripping fabric naturally, long tail feathers extending downward, glossy plumage catching light, slight wing adjustment for balance, natural weight distribution, soft shadow beneath bird."],
-            ["examples/painter.mp4",  "Change the hair color to platinum blonde with natural highlights, subtle root shadowing, silky texture, gentle waves, soft shine, dimensional tones, strand definition, natural movement, professional color treatment, salon-quality finish, light-catching shimmer, varied blonde shades from honey to ash, realistic color gradation, healthy glossy appearance, volumetric lighting interaction."],
-        ],
-        inputs=[video_input, prompt],
-        outputs=video_output,
-        fn=process_video,
-        cache_examples="lazy",
-    )
-
+    examples=[
+        ["examples/man_walking.mp4", "make the man into an alien"],
+        ["examples/leopard.mp4",  "make the leopard into a lion"],
+        ["examples/woman.mp4", "make the woman's coat blue"],
+    ],
+    inputs=[video_input, prompt],
+    outputs=video_output,
+    fn=process_video,
+    cache_examples="lazy",
+)
+    
     # Event handlers
     generate_btn.click(
         fn=process_video,
@@ -232,4 +242,4 @@ with gr.Blocks(title="Lucy Edit - Video Editing with Text", css=css) as demo:
     )
 
 if __name__ == "__main__":
-    demo.launch(share=True)
+    demo.launch(share=True)

examples/neon.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:e62dc89173956d4d22defac1d24f0835c8d8d0490f7969dd13b38429d69165ca
-size 3182890

examples/painter.mp4

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:37b385068db4c25d8f29104580b3267a4fafc2afbd4235f57a206a587a16e56f
-size 2941421