Update app.py

app.py

@@ -15,6 +15,7 @@ import cv2
 import tempfile
 import os
 
+# Load models and configurations
 config_file = hf_hub_download(
     "xinsir/controlnet-union-sdxl-1.0",
     filename="config_promax.json",
@@ -169,39 +170,35 @@ def infer(image, width=1024, height=1024, overlap_width=18, num_inference_steps=
 
     yield background, cnet_image
 
-def clear_result():
-    """Clears the result ImageSlider."""
-    return gr.update(value=None)
-
-def preload_presets(target_ratio, ui_width, ui_height):
-    """Updates the width and height sliders based on the selected aspect ratio."""
-    if target_ratio == "9:16":
-        changed_width = 720
-        changed_height = 1280
-        return changed_width, changed_height, gr.update(open=False)
-    elif target_ratio == "16:9":
-        changed_width = 1280
-        changed_height = 720
-        return changed_width, changed_height, gr.update(open=False)
-    elif target_ratio == "1:1":
-        changed_width = 1024
-        changed_height = 1024
-        return changed_width, changed_height, gr.update(open=False)
-    elif target_ratio == "Custom":
-        return ui_width, ui_height, gr.update(open=True)
-
-def select_the_right_preset(user_width, user_height):
-    if user_width == 720 and user_height == 1280:
-        return "9:16"
-    elif user_width == 1280 and user_height == 720:
-        return "16:9"
-    elif user_width == 1024 and user_height == 1024:
-        return "1:1"
-    else:
-        return "Custom"
+def interpolate_frames(frame1, frame2, num_intermediate_frames):
+    """
+    Interpolate between two frames by gradually zooming out from frame2 to frame1.
+    """
+    frame1 = np.array(frame1)
+    frame2 = np.array(frame2)
+    h, w = frame1.shape[:2]
+    frames = []
+
+    for i in range(num_intermediate_frames + 2):
+        progress = i / (num_intermediate_frames + 1)
+        # Calculate the size of the inner frame
+        inner_h = int(h * (1 - progress))
+        inner_w = int(w * (1 - progress))
+        
+        # Crop the center of frame2
+        start_y = (h - inner_h) // 2
+        start_x = (w - inner_w) // 2
+        cropped = frame2[start_y:start_y+inner_h, start_x:start_x+inner_w]
+        
+        # Resize the cropped image to full size
+        interpolated = Image.fromarray(cropped).resize((w, h), Image.LANCZOS)
+        interpolated = np.array(interpolated)
+        
+        # Blend with frame1
+        blended = (1 - progress) * frame1 + progress * interpolated
+        frames.append(Image.fromarray(blended.astype(np.uint8)))
 
-def toggle_custom_resize_slider(resize_option):
-    return gr.update(visible=(resize_option == "Custom"))
+    return frames
 
 def create_video_from_images(image_list, fps=4):
     if not image_list:
@@ -222,28 +219,68 @@ def create_video_from_images(image_list, fps=4):
     return video_path
 
 @spaces.GPU(duration=70)
-def loop_outpainting(image, width=1024, height=1024, overlap_width=18, num_inference_steps=8, resize_option="custom", custom_resize_size=768, prompt_input=None, alignment="Middle", num_iterations=18, fps=6, progress=gr.Progress()):
+def loop_outpainting(image, width=1024, height=1024, overlap_width=18, num_inference_steps=8, 
+                     resize_option="custom", custom_resize_size=768, prompt_input=None, 
+                     alignment="Middle", num_iterations=18, fps=6, num_interpolation_frames=5,
+                     progress=gr.Progress()):
     image_list = [image]
     current_image = image
 
     for _ in progress.tqdm(range(num_iterations), desc="Generating frames"):
         # Generate new image
-        for step_result in infer(current_image, width, height, overlap_width, num_inference_steps, resize_option, custom_resize_size, prompt_input, alignment):
+        for step_result in infer(current_image, width, height, overlap_width, num_inference_steps, 
+                                 resize_option, custom_resize_size, prompt_input, alignment):
             pass  # Process all steps
         
         new_image = step_result[1]  # Get the final image from the last step
-        image_list.append(new_image)
+        
+        # Interpolate between current_image and new_image
+        interpolated_frames = interpolate_frames(current_image, new_image, num_interpolation_frames)
+        image_list.extend(interpolated_frames)
         
         # Use new image as input for next iteration
         current_image = new_image
 
-    reverse_image_list = image_list[::-1]
     # Create video from image list
-    video_path = create_video_from_images(reverse_image_list, fps)
+    video_path = create_video_from_images(image_list, fps)
     return video_path
 
 loop_outpainting.zerogpu = True
 
+def clear_result():
+    """Clears the result ImageSlider."""
+    return gr.update(value=None)
+
+def preload_presets(target_ratio, ui_width, ui_height):
+    """Updates the width and height sliders based on the selected aspect ratio."""
+    if target_ratio == "9:16":
+        changed_width = 720
+        changed_height = 1280
+        return changed_width, changed_height, gr.update(open=False)
+    elif target_ratio == "16:9":
+        changed_width = 1280
+        changed_height = 720
+        return changed_width, changed_height, gr.update(open=False)
+    elif target_ratio == "1:1":
+        changed_width = 1024
+        changed_height = 1024
+        return changed_width, changed_height, gr.update(open=False)
+    elif target_ratio == "Custom":
+        return ui_width, ui_height, gr.update(open=True)
+
+def select_the_right_preset(user_width, user_height):
+    if user_width == 720 and user_height == 1280:
+        return "9:16"
+    elif user_width == 1280 and user_height == 720:
+        return "16:9"
+    elif user_width == 1024 and user_height == 1024:
+        return "1:1"
+    else:
+        return "Custom"
+
+def toggle_custom_resize_slider(resize_option):
+    return gr.update(visible=(resize_option == "Custom"))
+
 css = """
 .gradio-container {
     width: 1200px !important;
@@ -326,6 +363,8 @@ with gr.Blocks(css=css) as demo:
                         with gr.Row():
                             num_iterations = gr.Slider(label="Number of iterations", minimum=2, maximum=24, step=1, value=18)
                             fps = gr.Slider(label="fps", minimum=1, maximum=24, value=8)
+                        with gr.Row():
+                            num_interpolation_frames = gr.Slider(label="Interpolation frames", minimum=0, maximum=10, step=1, value=5)
 
             with gr.Column():
                 result = ImageSlider(
@@ -335,6 +374,7 @@ with gr.Blocks(css=css) as demo:
                 )
                 use_as_input_button = gr.Button("Use as Input Image", visible=False)
                 video_output = gr.Video(label="Outpainting Video")
+
     gr.Examples(
         examples=["hide.png", "disaster.png"],
         fn=loop_outpainting,
@@ -342,6 +382,7 @@ with gr.Blocks(css=css) as demo:
         outputs=video_output,
         cache_examples="lazy"
     )
+
     def use_output_as_input(output_image):
         """Sets the generated output as the new input image."""
         return gr.update(value=output_image[1])
@@ -413,7 +454,8 @@ with gr.Blocks(css=css) as demo:
     loop_button.click(
         fn=loop_outpainting,
         inputs=[input_image, width_slider, height_slider, overlap_width, num_inference_steps,
-                resize_option, custom_resize_size, prompt_input, alignment_dropdown, num_iterations, fps],
+                resize_option, custom_resize_size, prompt_input, alignment_dropdown, 
+                num_iterations, fps, num_interpolation_frames],
         outputs=video_output,
     )