Update app.py

app.py

@@ -54,20 +54,26 @@ if __name__ == "__main__":
         with gr.Row():
             with gr.Column(variant='panel'):
 
-                gr.Markdown('## Select input video and audio', sanitize_html=False)
-                # Define samples
+                gr.Markdown('## Select or Upload input video and audio', sanitize_html=False)
+                # Define preview slots
                 sample_video = gr.Video(interactive=False, label="Input Video")
                 sample_audio = gr.Audio(interactive=False, label="Input Audio")
 
+                # New upload inputs
+                video_upload = gr.Video(source="upload", type="filepath", label="Upload Video")
+                audio_upload = gr.Audio(source="upload", type="filepath", label="Upload Audio")
+
                 # Define radio inputs
-                video_selection = gr.components.Radio(video_label_dict,
-                                                      type='value', label="Select an input video:")
-                audio_selection = gr.components.Radio(audio_label_dict,
-                                                      type='value', label="Select an input audio:")
+                video_selection = gr.Radio(video_label_dict,
+                                           type='value', label="Select an input video:")
+                audio_selection = gr.Radio(audio_label_dict,
+                                           type='value', label="Select an input audio:")
+
                 # Define button inputs
                 with gr.Row(equal_height=True):
                     generate_original_button = gr.Button(value="Generate with Original Model", variant="primary")
                     generate_compressed_button = gr.Button(value="Generate with Compressed Model", variant="primary")
+
             with gr.Column(variant='panel'):
                 # Define original model output components
                 gr.Markdown('## Original Wav2Lip')
@@ -87,19 +93,34 @@ if __name__ == "__main__":
                         compressed_model_fps = gr.Textbox(value="", label="FPS")
                     compressed_model_params = gr.Textbox(value=servicer.params['nota_wav2lip'], label="# Parameters")
 
-        # Switch video and audio samples when selecting the raido button
+        # Switch video and audio samples when selecting the radio button
         video_selection.change(fn=servicer.switch_video_samples, inputs=video_selection, outputs=sample_video)
         audio_selection.change(fn=servicer.switch_audio_samples, inputs=audio_selection, outputs=sample_audio)
 
+        # Update preview when uploading
+        video_upload.change(fn=lambda x: x, inputs=video_upload, outputs=sample_video)
+        audio_upload.change(fn=lambda x: x, inputs=audio_upload, outputs=sample_audio)
+
+        # Helper: decide whether to use uploaded or selected
+        def resolve_inputs(video_choice, audio_choice, video_file, audio_file):
+            video_path = video_file if video_file else video_label_dict.get(video_choice)
+            audio_path = audio_file if audio_file else audio_label_dict.get(audio_choice)
+            return video_path, audio_path
+
         # Click the generate button for original model
-        generate_original_button.click(servicer.generate_original_model,
-                                       inputs=[video_selection, audio_selection],
-                                       outputs=[original_model_output, original_model_inference_time, original_model_fps])
+        generate_original_button.click(
+            fn=lambda v, a, vu, au: servicer.generate_original_model(*resolve_inputs(v, a, vu, au)),
+            inputs=[video_selection, audio_selection, video_upload, audio_upload],
+            outputs=[original_model_output, original_model_inference_time, original_model_fps]
+        )
+
         # Click the generate button for compressed model
-        generate_compressed_button.click(servicer.generate_compressed_model,
-                                         inputs=[video_selection, audio_selection],
-                                         outputs=[compressed_model_output, compressed_model_inference_time, compressed_model_fps])
+        generate_compressed_button.click(
+            fn=lambda v, a, vu, au: servicer.generate_compressed_model(*resolve_inputs(v, a, vu, au)),
+            inputs=[video_selection, audio_selection, video_upload, audio_upload],
+            outputs=[compressed_model_output, compressed_model_inference_time, compressed_model_fps]
+        )
 
         gr.Markdown(Path('docs/footer.md').read_text())
 
-    demo.queue().launch()
+    demo.queue().launch()