Update app.py

app.py

@@ -89,15 +89,16 @@ def infer(video_input, manual_caption, duration_in, seed, caption_output):
     duration = video_in.duration
     if duration > 5:
         video_in = video_in.subclip(0,5)
-    
+
+    new_audio = AudioFileClip("sound.mp3")
     # Make the audio the same length as the video
-    audio = audio.set_duration(video_in.duration)
+    new_audio = new_audio.set_duration(video_in.duration)
     
     # Combine the audio and video
-    result = video_in.set_audio(audio)
+    result = video_in.set_audio(new_audio)
     
     # Save the result
-    result.write_videofile("result.mp4")
+    result.write_videofile("result.mp4", codec='libx264', audio_codec='aac')
 
     
     #return cap, sound[1], gr.Textbox.update(placeholder=f"{ph_update}{ph_message}"), gr.Group.update(visible=True)