Implementation:Facebookresearch Audiocraft MusicGenStyleApp

Knowledge Sources	Facebookresearch_Audiocraft
Domains	Demo, Music_Generation
Last Updated	2026-02-14 01:00 GMT

Overview

Concrete tool for running an interactive Gradio web demo of the MusicGen-Style model for text-and-style-conditioned music generation.

Description

MusicGenStyleApp is a Gradio web application that provides an interactive interface for the MusicGen-Style model. It supports both text-only and text-plus-style-audio conditioned generation, allowing users to provide an audio clip as a style reference. The demo includes configurable generation parameters such as duration, temperature, top-k/top-p sampling, and classifier-free guidance coefficient.

Usage

Run this application to demonstrate MusicGen-Style's text-and-audio-conditioned music generation through a web browser interface.

Code Reference

Source Location

• Repository: Facebookresearch_Audiocraft
• File: demos/musicgen_style_app.py
• Lines: 1-380

Signature

def predict_full(model, text, melody, duration, topk, topp, temperature, cfg_coef):
    """Generate music from text and optional style audio reference."""

Import

# Run as a standalone script
python demos/musicgen_style_app.py

I/O Contract

Inputs

Name	Type	Required	Description
text	str	Yes	Text description for music generation
melody	tuple	No	Style audio reference (sample_rate, waveform)
duration	float	No	Generation duration in seconds
model	str	Yes	Model identifier to use

Outputs

Name	Type	Description
audio	tuple(int, numpy.ndarray)	Generated audio as (sample_rate, waveform)

Usage Examples

# Launch the demo
python demos/musicgen_style_app.py --share

# Or programmatically:
from audiocraft.models import MusicGen
model = MusicGen.get_pretrained('facebook/musicgen-style')
model.set_generation_params(duration=10)
wav = model.generate(['upbeat electronic dance music'])

Related Pages

• Principle:Facebookresearch_Audiocraft_Generation_Parameter_Configuration