Update app.py

app.py

@@ -102,5 +102,5 @@ def generate_audio(video_file, seed_val, cfg_scale, num_steps, mode):
     ffmpeg.output(input_v, input_a, output_video, vcodec="libx264", acodec="aac", strict="experimental").run(overwrite_output=True, quiet=True)
     return output_video, audio_path
 
-demo = gr.Interface(fn=generate_audio, inputs=[gr.Video(label="Input Video"), gr.Number(label="Seed", value=0, precision=0), gr.Slider(label="CFG Scale", minimum=1, maximum=15, value=8, step=0.5), gr.Slider(label="Sampling Steps", minimum=10, maximum=50, value=25, step=1), gr.Radio(label="Sampling Mode", choices=["sde", "ode"], value="sde")], outputs=[gr.Video(label="Output Video with Audio"), gr.Audio(label="Generated Audio")], title="TARO: Video-to-Audio Synthesis (ICCV 2025)", description="Upload a video and generate synchronized audio using TARO. 8.19 seconds is the optimal duration")
+demo = gr.Interface(fn=generate_audio, inputs=[gr.Video(label="Input Video"), gr.Number(label="Seed", value=0, precision=0), gr.Slider(label="CFG Scale", minimum=1, maximum=15, value=8, step=0.5), gr.Slider(label="Sampling Steps", minimum=10, maximum=50, value=25, step=1), gr.Radio(label="Sampling Mode", choices=["sde", "ode"], value="sde")], outputs=[gr.Video(label="Output Video with Audio"), gr.Audio(label="Generated Audio")], title="TARO: Video-to-Audio Synthesis (ICCV 2025)", description="Upload a video and generate synchronized audio using TARO. Optimal duration is as close to 8 seconds as possible.")
 demo.queue().launch()