Update app.py

app.py

@@ -49,18 +49,18 @@ with gr.Blocks(title="🎹 MIDI-AudioLDM", theme=gr.themes.Base(text_size=gr.the
  """)
  with gr.Row():
  with gr.Column(variant='panel'):
- midi = gr.File(label="midi file", file_types=[".mid"])
- prompt = gr.Textbox(label="prompt", info="Enter a descriptive text prompt to guide the audio generation.")
+ midi = gr.File(label="Midi file", file_types=[".mid"])
+ prompt = gr.Textbox(label="Prompt", info="Enter a descriptive text prompt to guide the audio generation.")
  with gr.Column(variant='panel'):
- audio = gr.Audio(label="audio")
+ audio = gr.Audio(label="Audio")
  with gr.Accordion("Advanced Settings", open=False):
- duration = gr.Slider(0, 30, value=5, step=5, label="duration (seconds)", info="Modify the duration of the output audio file.")
- inf = gr.Slider(0, 50, value=20, step=0.1, label="inference steps", info="Edit the number of denoising steps. A larger number usually leads to higher quality but slower results.")
- guidance_scale = gr.Slider(0, 4, value=2.5, step=0.5, label="guidance scale", info="Modify the guidance scale. The higher the value the more linked the generated audio to the text prompt, sometimes at the expense of lower quality.")
- neg_prompt = gr.Textbox(label="negative prompt", info="Optionally enter a negative text prompt not to guide the audio generation.")
- seed = gr.Number(value=25, label="random seed", info="Change the random seed for a different generation result.")
- cond = gr.Slider(0.0, 1.0, value=1.0, step=0.1, label="conditioning scale", info="Enter a value between 0 and 1. The larger the more it will take the conditioning into account. Lower values are recommended for more creative prompts.")
- guess = gr.Checkbox(label="guess mode", info="If selected, the model will try to recognize the content of the MIDI without the need of a text prompt.")
+ duration = gr.Slider(0, 30, value=5, step=5, label="Duration (seconds)", info="Modify the duration of the output audio file.")
+ inf = gr.Slider(0, 50, value=20, step=0.1, label="Inference steps", info="Edit the number of denoising steps. A larger number usually leads to higher quality but slower results.")
+ guidance_scale = gr.Slider(0, 4, value=2.5, step=0.5, label="Guidance scale", info="Modify the guidance scale. The higher the value the more linked the generated audio to the text prompt, sometimes at the expense of lower quality.")
+ neg_prompt = gr.Textbox(label="Negative prompt", info="Optionally enter a negative text prompt not to guide the audio generation.")
+ seed = gr.Number(value=25, label="Random seed", info="Change the random seed for a different generation result.")
+ cond = gr.Slider(0.0, 1.0, value=1.0, step=0.1, label="Conditioning scale", info="Enter a value between 0 and 1. The larger the more it will take the conditioning into account. Lower values are recommended for more creative prompts.")
+ guess = gr.Checkbox(label="Guess mode", info="If selected, the model will try to recognize the content of the MIDI without the need of a text prompt.")
  btn = gr.Button("Generate")
  btn.click(predict, inputs=[midi, prompt, neg_prompt, duration, seed, cond, inf, guidance_scale, guess], outputs=[audio])
  # gr.Examples(examples=[["S00.mid", "piano", "", 10, 25, 1.0, 20, 2.5, False], ["S00.mid", "violin", "", 10, 25, 1.0, 20, 2.5, False], ["S00.mid", "woman singing", "", 10, 25, 0.8, 20, 2.5, False]], inputs=[midi, prompt, neg_prompt, duration, seed, cond, inf, guidance_scale, guess], fn=predict, outputs=audio, cache_examples=True)