continuing continuation attempt 1

app.py

@@ -182,6 +182,58 @@ def generate_music(wav_filename, prompt_duration, musicgen_model, num_iterations
 
     return combined_audio_filename
 
+    def continue_music(input_audio_path, prompt_duration, musicgen_model, num_iterations, bpm):
+    # Load the audio from the given file path
+    song, sr = torchaudio.load(input_audio_path)
+    song = song.to(device)
+    
+    # Calculate the slice from the end of the song based on prompt_duration
+    num_samples = int(prompt_duration * sr)
+    if song.shape[-1] < num_samples:
+        raise ValueError("The prompt_duration is longer than the audio length.")
+    start_idx = song.shape[-1] - num_samples
+    prompt_waveform = song[..., start_idx:]
+
+    # Prepare the audio slice for generation
+    prompt_waveform = preprocess_audio(prompt_waveform)
+    
+    # Load the model and set generation parameters as before
+    model_continue = MusicGen.get_pretrained(musicgen_model.split(" ")[0])
+    model_continue.set_generation_params(
+        use_sampling=True,
+        top_k=250,
+        top_p=0.0,
+        temperature=1.0,
+        duration=calculate_duration(bpm),
+        cfg_coef=3
+    )
+    
+    all_audio_files = []
+    for i in range(num_iterations):
+        output = model_continue.generate_continuation(prompt_waveform, prompt_sample_rate=sr, progress=True)
+        output = output.cpu()  # Ensure the output is on CPU for further processing
+        if len(output.size()) > 2:
+            output = output.squeeze()
+
+        filename_without_extension = f'continue_{i}'
+        filename_with_extension = f'{filename_without_extension}.wav'
+        audio_write(filename_with_extension, output, model_continue.sample_rate, strategy="loudness", loudness_compressor=True)
+        all_audio_files.append(filename_with_extension)
+
+    # Combine all audio files as before
+    combined_audio = AudioSegment.empty()
+    for filename in all_audio_files:
+        combined_audio += AudioSegment.from_wav(filename)
+
+    combined_audio_filename = f"combined_audio_{random.randint(1, 10000)}.mp3"
+    combined_audio.export(combined_audio_filename, format="mp3")
+
+    # Clean up temporary files
+    for filename in all_audio_files:
+        os.remove(filename)
+
+    return combined_audio_filename
+
 # Define the expandable sections
 musiclang_blurb = """
 ## musiclang
@@ -234,12 +286,15 @@ with gr.Blocks() as iface:
                 "thepatch/bleeps-medium (medium)",
                 "thepatch/hoenn_lofi (large)"
             ], value="thepatch/vanya_ai_dnb_0.1 (small)")
-            num_iterations = gr.Slider(label="Number of Iterations", minimum=1, maximum=10, step=1, value=3)
+            num_iterations = gr.Slider(label="Number of Iterations", minimum=1, maximum=3, step=1, value=3)
             generate_music_button = gr.Button("Generate Music")
             output_audio = gr.Audio(label="Generated Music")
+            continue_button = gr.Button("Continue Generating Music")
+            continue_output_audio = gr.Audio(label="Continued Music Output")
 
     # Connecting the components
     generate_midi_button.click(generate_midi, inputs=[seed, use_chords, chord_progression, bpm], outputs=[midi_audio])
     generate_music_button.click(generate_music, inputs=[midi_audio, prompt_duration, musicgen_model, num_iterations, bpm], outputs=[output_audio])
+    continue_button.click(continue_music, inputs=[output_audio, prompt_duration, musicgen_model, num_iterations, bpm], outputs=continue_output_audio)
 
 iface.launch()