Update app.py

app.py

@@ -1,33 +1,38 @@
-# Import necessary libraries
-from audiocraft.models import MusicGen  # Import the MusicGen model for music generation
-from audiocraft.data.audio import audio_write # Import audio_write function for saving audio
-import gradio as gr  # Import Gradio for creating a web interface
-import torch  # Import PyTorch for deep learning operations
+import tempfile
+from audiocraft.models import MusicGen
+from audiocraft.data.audio import audio_write
+import gradio as gr
+import torch
+import tempfile
+import uuid
+import os
+from scipy.io.wavfile import write
 
-# Load the pretrained MusicGen model (small)
+# Load the model and set parameters
 model = MusicGen.get_pretrained("facebook/musicgen-small")
+model.set_generation_params(duration=8)
 
-# Set generation parameters for the model (e.g., duration of 8 seconds)
-model.set_generation_params(duration=10)
-
-# Function to generate music from text descriptions
 def generate_music(description):
-    # Generate music based on the provided description
+    # Generate audio
     wav = model.generate([description])
-    
-    # Convert the audio waveform to a format that Gradio can display
-    audio_data = wav[0].cpu().numpy().tobytes()
-
-    return audio_data  # Return the audio data
+    audio_array = wav.cpu().numpy().squeeze()
+    sample_rate = model.sample_rate
+    file_id = uuid.uuid1()
+    file_path = os.path.join(
+        tempfile.gettempdir(),
+        f'{file_id}.wav'
+    )
+    write(file_path, rate=sample_rate, data=audio_array)
+    return file_path
 
-# Define a Gradio interface
+# Define Gradio interface with temporary file output
 iface = gr.Interface(
-    fn=generate_music,  # Use the generate_music function for processing input
-    inputs="text",  # Accept text input from the user
-    outputs=gr.Audio(type="numpy", label="Audio"),  # Display the generated audio as output
-    title="Music Generation from Descriptions",  # Set the title of the web interface
-    description="Generate music based on descriptions.",  # Provide a description
-    live=False  # Set to False if you don't want real-time updates (for beginner-friendly interaction)
+    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(debug=True)