Update app.py

app.py

@@ -64,9 +64,13 @@ with gr.Blocks(title="🎹 MIDI-AudioLDM", theme=gr.themes.Base(text_size=gr.the
             """)
     with gr.Column(variant='panel'):
         midi = gr.File(label="midi file", file_types=[".mid"])
-        midi_synth = gr.Audio(label="synthesized midi")
-        midi.upload(synthesize, midi, midi_synth)
         prompt = gr.Textbox(label="prompt", info="Enter a descriptive text prompt to guide the audio generation.")
+        with gr.Row():
+            with gr.Column():
+                midi_synth = gr.Audio(label="synthesized midi")
+                midi.upload(synthesize, midi, midi_synth)
+            with gr.Column():
+                audio = gr.Audio(label="generated audio")
         with gr.Accordion("Advanced Settings", open=False):
             duration = gr.Slider(0, 30, value=10, step=2.5, label="duration", info="Modify the duration in seconds of the output audio file.")
             inf = gr.Slider(0, 50, value=20, step=1, label="inference steps", info="Edit the number of denoising steps. A larger number usually leads to higher quality but slower results.")
@@ -75,8 +79,6 @@ with gr.Blocks(title="🎹 MIDI-AudioLDM", theme=gr.themes.Base(text_size=gr.the
             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="Choose 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="Optionally select guess mode. If so, the model will try to recognize the content of the MIDI without the need of a text prompt.")
-    with gr.Column(variant='panel'):
-        audio = gr.Audio(label="generated audio")    
     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, studio recording", "noise", 10, 25, 1.0, 20, 2.5, False], ["S00.mid", "jazz band, clean", "noise", 10, 25, 0.8, 20, 2.5, False], ["S00.mid", "choir", "noise, percussion", 10, 25, 0.7, 20, 2.5, False]], inputs=[midi, prompt, neg_prompt, duration, seed, cond, inf, guidance_scale, guess], fn=run_example, outputs=[midi_synth, audio], cache_examples=True)