update

app.py

@@ -223,6 +223,15 @@ def perform_anyv2v(
     return output_path
 
 
+
+def get_first_frame_as_pil(video_path):
+    with VideoFileClip(video_path) as clip:
+        # Extract the first frame (at t=0) as an array
+        first_frame_array = clip.get_frame(0)
+        # Convert the numpy array to a PIL Image
+        first_frame_image = Image.fromarray(first_frame_array)
+        return first_frame_image
+        
 def btn_preprocess_video_fn(video_path, width, height, start_time, end_time, center_crop, x_offset, y_offset, longest_to_width):
     def check_video(video_path):
         with VideoFileClip(video_path) as clip:
@@ -231,14 +240,6 @@ def btn_preprocess_video_fn(video_path, width, height, start_time, end_time, cen
             else:
                 return False
 
-    def get_first_frame_as_pil(video_path):
-        with VideoFileClip(video_path) as clip:
-            # Extract the first frame (at t=0) as an array
-            first_frame_array = clip.get_frame(0)
-            # Convert the numpy array to a PIL Image
-            first_frame_image = Image.fromarray(first_frame_array)
-            return first_frame_image
-        
     if check_video(video_path) == False:
         processed_video_path = crop_and_resize_video(input_video_path=video_path, 
                                                     output_folder=TEMP_DIR,
@@ -251,11 +252,10 @@ def btn_preprocess_video_fn(video_path, width, height, start_time, end_time, cen
                                                     x_offset=x_offset, 
                                                     y_offset=y_offset, 
                                                     longest_to_width=longest_to_width)
-        frame = get_first_frame_as_pil(processed_video_path)
-        return processed_video_path, frame
+
+        return processed_video_path
     else:
-        frame = get_first_frame_as_pil(video_path)
-        return video_path, frame
+        return video_path
 
 def btn_image_edit_fn(video_path, instruct_prompt, ie_force_512, ie_seed, ie_neg_prompt):
     """
@@ -268,7 +268,7 @@ def btn_image_edit_fn(video_path, instruct_prompt, ie_force_512, ie_seed, ie_neg
         ie_seed = int.from_bytes(os.urandom(2), "big")
     print(f"Using seed: {ie_seed}")
 
-    edited_image_path = perform_edit(video_path=video_path, 
+    edited_image_path = Image_Editor.perform_edit(video_path=video_path, 
                                              prompt=instruct_prompt,
                                              force_512=ie_force_512,
                                              seed=ie_seed,
@@ -293,7 +293,7 @@ def btn_infer_fn(video_path,
         seed = int.from_bytes(os.urandom(2), "big")
     print(f"Using seed: {seed}")
 
-    result_video_path = perform_anyv2v(video_path=video_path,
+    result_video_path = AnyV2V_Editor.perform_anyv2v(video_path=video_path,
                                                         video_prompt=video_prompt,
                                                         video_negative_prompt=video_negative_prompt,
                                                         edited_first_frame_path=edited_first_frame_path,
@@ -391,7 +391,7 @@ with gr.Blocks() as demo:
     btn_pv.click(
         btn_preprocess_video_fn,
         inputs=[video_raw, pv_width, pv_height, pv_start_time, pv_end_time, pv_center_crop, pv_x_offset, pv_y_offset, pv_longest_to_width],
-        outputs=[video_input, src_first_frame]
+        outputs=video_input
     )
 
     btn_image_edit.click(
@@ -416,6 +416,9 @@ with gr.Blocks() as demo:
                 av_seed],
         outputs=video_output
     )
+
+    video_input.change(fn=get_first_frame_as_pil, inputs=video_input, outputs=src_first_frame)
+
 #=====================================
 
 # Minimizing usage of GPU Resources

gradio_demo.py

@@ -223,6 +223,15 @@ def perform_anyv2v(
     return output_path
 
 
+
+def get_first_frame_as_pil(video_path):
+    with VideoFileClip(video_path) as clip:
+        # Extract the first frame (at t=0) as an array
+        first_frame_array = clip.get_frame(0)
+        # Convert the numpy array to a PIL Image
+        first_frame_image = Image.fromarray(first_frame_array)
+        return first_frame_image
+        
 def btn_preprocess_video_fn(video_path, width, height, start_time, end_time, center_crop, x_offset, y_offset, longest_to_width):
     def check_video(video_path):
         with VideoFileClip(video_path) as clip:
@@ -231,14 +240,6 @@ def btn_preprocess_video_fn(video_path, width, height, start_time, end_time, cen
             else:
                 return False
 
-    def get_first_frame_as_pil(video_path):
-        with VideoFileClip(video_path) as clip:
-            # Extract the first frame (at t=0) as an array
-            first_frame_array = clip.get_frame(0)
-            # Convert the numpy array to a PIL Image
-            first_frame_image = Image.fromarray(first_frame_array)
-            return first_frame_image
-        
     if check_video(video_path) == False:
         processed_video_path = crop_and_resize_video(input_video_path=video_path, 
                                                     output_folder=TEMP_DIR,
@@ -251,11 +252,10 @@ def btn_preprocess_video_fn(video_path, width, height, start_time, end_time, cen
                                                     x_offset=x_offset, 
                                                     y_offset=y_offset, 
                                                     longest_to_width=longest_to_width)
-        frame = get_first_frame_as_pil(processed_video_path)
-        return processed_video_path, frame
+
+        return processed_video_path
     else:
-        frame = get_first_frame_as_pil(video_path)
-        return video_path, frame
+        return video_path
 
 def btn_image_edit_fn(video_path, instruct_prompt, ie_force_512, ie_seed, ie_neg_prompt):
     """
@@ -268,7 +268,7 @@ def btn_image_edit_fn(video_path, instruct_prompt, ie_force_512, ie_seed, ie_neg
         ie_seed = int.from_bytes(os.urandom(2), "big")
     print(f"Using seed: {ie_seed}")
 
-    edited_image_path = perform_edit(video_path=video_path, 
+    edited_image_path = Image_Editor.perform_edit(video_path=video_path, 
                                              prompt=instruct_prompt,
                                              force_512=ie_force_512,
                                              seed=ie_seed,
@@ -293,7 +293,7 @@ def btn_infer_fn(video_path,
         seed = int.from_bytes(os.urandom(2), "big")
     print(f"Using seed: {seed}")
 
-    result_video_path = perform_anyv2v(video_path=video_path,
+    result_video_path = AnyV2V_Editor.perform_anyv2v(video_path=video_path,
                                                         video_prompt=video_prompt,
                                                         video_negative_prompt=video_negative_prompt,
                                                         edited_first_frame_path=edited_first_frame_path,
@@ -391,7 +391,7 @@ with gr.Blocks() as demo:
     btn_pv.click(
         btn_preprocess_video_fn,
         inputs=[video_raw, pv_width, pv_height, pv_start_time, pv_end_time, pv_center_crop, pv_x_offset, pv_y_offset, pv_longest_to_width],
-        outputs=[video_input, src_first_frame]
+        outputs=video_input
     )
 
     btn_image_edit.click(
@@ -416,6 +416,9 @@ with gr.Blocks() as demo:
                 av_seed],
         outputs=video_output
     )
+
+    video_input.change(fn=get_first_frame_as_pil, inputs=video_input, outputs=src_first_frame)
+
 #=====================================
 
 # Minimizing usage of GPU Resources