Update app.py

app.py

@@ -74,11 +74,26 @@ def infer(video_input, manual_caption, duration_in, seed, caption_output):
     # AudioLDM loaded demo returns a video, so we only keep the audio
     video = VideoFileClip(sound)
     audio = video.audio
-    audio.write_audiofile("sound.mp3")
+    #audio.write_audiofile("sound.mp3")
+
+    # Then we put the audio to the original video
+    
+    # Load the input video file
+    video_in = VideoFileClip(video_input)
+    
+    # Make the audio the same length as the video
+    audio = audio.set_duration(video.duration)
+    
+    # Combine the audio and video
+    result = video_in.set_audio(audio)
+    
+    # Save the result
+    result.write_videofile("result.mp4")
+
     
     #return cap, sound[1], gr.Textbox.update(placeholder=f"{ph_update}{ph_message}"), gr.Group.update(visible=True)
     #return cap, "sound.mp3", gr.Group.update(visible=True)
-    return cap, "sound.mp3"
+    return cap, "result.mp4"
 
 title = """
     <div style="text-align: center; max-width: 700px; margin: 0 auto;">
@@ -139,13 +154,13 @@ with gr.Blocks(css="style.css") as demo:
         input_vid = gr.Video(source="upload", type="filepath", elem_id="input-vid")
         
         with gr.Column():
-            manual_cap = gr.Textbox(label="Manual Image description (optional)", lines=3, placeholder=ph_message)
+            manual_cap = gr.Textbox(label="Manual Video description (optional)", lines=3, placeholder=ph_message)
             with gr.Row():
                 duration_in = gr.Slider(interactive=False, minimum=5, maximum=10, step=5, value=5, label="Duration")
                 seed_in = gr.Slider(label="Seed", value=440, minimum=45, maximum=10000, step=1)
         
         caption_output = gr.Textbox(label="Caption", visible=False, elem_id="text-caption")
-        sound_output = gr.Audio(label="Result", elem_id="sound-output")
+        video_output = gr.Video(label="Result", elem_id="video-output")
         #debug = gr.Textbox()
         generate = gr.Button("Generate SFX from Video")
 
@@ -156,12 +171,12 @@ with gr.Blocks(css="style.css") as demo:
 
         gr.HTML(article)
 
-    change_out = [manual_cap, caption_output, sound_output]
+    change_out = [manual_cap, caption_output, video_output]
     input_vid.change(input_changes, input_vid, change_out, queue=False)
     
 
     
-    generate.click(infer, inputs=[input_vid, manual_cap, duration_in, seed_in, caption_output], outputs=[caption_output, sound_output], api_name="v2fx")
+    generate.click(infer, inputs=[input_vid, manual_cap, duration_in, seed_in, caption_output], outputs=[caption_output, video_output], api_name="v2fx")
     #share_button.click(None, [], [], _js=share_js)
 
 demo.queue(max_size=32).launch(debug=True)