Update app.py

app.py

@@ -1,41 +1,44 @@
-import gradio as gr
-from transformers import MusicGenForConditionalGeneration, MusicGenProcessor
-import torch
-import soundfile as sf  
+import tempfile 
+from audiocraft.models import MusicGen 
+from audiocraft.data.audio import audio_write 
+import gradio as gr 
+import torch 
+import uuid 
+import os 
+from scipy.io.wavfile import write 
 
-from transformers import AutoModel, AutoProcessor
 
-# Use AutoModel and AutoProcessor to automatically select the correct model and processor
-model_name = "facebook/musicgen-small"
-model = AutoModel.from_pretrained(model_name)
-processor = AutoProcessor.from_pretrained(model_name)
+model = MusicGen.get_pretrained("facebook/musicgen-small") 
+model.set_generation_params(duration=5) 
 
-# Define a function to generate audio from text
-def generate_audio(text):
-    # Tokenize the input text
-    inputs = processor(text, return_tensors="pt")
+def generate_music(description):
+wav = model.generate([description]) 
+audio_array = wav.cpu().numpy().squeeze() 
+sample_rate = model.sample_rate 
 
-    # Generate audio (samples)
-    with torch.no_grad():
-        generated_audio = model.generate(**inputs)
 
-    # Convert tensor to numpy and save as a WAV file
-    audio_path = "/tmp/generated_audio.wav"
-    audio_data = generated_audio[0].cpu().numpy()  # Access the first sample
+file_id = uuid.uuid1()
 
-    # Save the generated audio
-    sf.write(audio_path, audio_data, 16000)  # Assuming a sample rate of 16kHz
+file_path = os.path.join(
+tempfile.gettempdir(), 
+f'{file_id}.wav' 
+)
+
+print(f"Temporary directory: {tempfile.gettempdir()}")
+print(f"File path: {file_path}")
+
+write(file_path, rate=sample_rate, data=audio_array)
 
-    return audio_path
+return file_path
 
-# Set up the Gradio interface
 iface = gr.Interface(
-    fn=generate_audio, 
-    inputs=gr.Textbox(label="Enter Text"), 
-    outputs=gr.Audio(type="file", label="Generated Audio"),
-    title="Text-to-Audio Chatbot",
-    description="Enter a text prompt and get a music clip generated by the MusicGen model."
+fn=generate_music, 
+inputs="text", 
+outputs=gr.components.Audio(type="filepath", label="Audio"), 
+title="Text to Audio Generation", 
+description="Generate audio based on text descriptions.", 
+live=False, 
 )
 
-iface.launch()
 
+iface.launch(debug=True)