MusicGen

MusicGen

Turns a text description into a short piece of music with the MusicGen model, in a browser demo built on the AudioCraft audio library. It suits musicians, developers and researchers exploring text-to-music.

Made by facebook on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: This Space is a demo of MusicGen, a text-to-music model, and its repository holds the AudioCraft library behind it. AudioCraft is a PyTorch library for audio generation research, with inference and training code for MusicGen and AudioGen. The demo runs on zero-a10g hardware at Hugging Face.

Who it suits: It suits people who want to hear what a text-to-music model produces and people who want to read or train the code behind it.

Industry | Who uses it | How they use it | Example
Music and sound | Musicians and sound designers | Try text-to-music generation in the browser | Describe a style in text and listen to the result
AI research | Audio researchers | Run and train audio generation models | Use the training code for MusicGen, EnCodec or Multi Band Diffusion
Software | Developers | Install the library in a Python project | Call the models from their own code
Education | Students | Study how audio models are built | Read the docs for each model

You can skip it if you want a polished music app, since this is a research library with a demo page.

Where to use it: You can use it in these places:

• A browser: open the Space's page on Hugging Face and nothing needs installing.
• Windows, macOS or Linux: run a copy of the demo with Python 3.10.
• Python projects: install the audiocraft package and call the models from code.

Use case | Who | What they get back
Text to music | Musicians and curious users | A generated piece of music from a text description
Text to sound | Sound designers | Sound effects from the AudioGen model
Model training | Researchers | Training pipelines for EnCodec, MusicGen and Multi Band Diffusion

How to set it up: You need these first for a local copy of the demo:

• Python 3.10 and pip
• ffmpeg, which the library README recommends

The online page needs none of this. For your own copy, follow the steps below.

1. Install the Hugging Face tools.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Install Gradio at the version the Space uses, then the other packages.
5. Start the demo from its main file.

pip install -U huggingface_hub
hf download facebook/MusicGen --repo-type space --local-dir MusicGen
cd MusicGen
pip install gradio==5.49.1
pip install -r requirements.txt
python demos/musicgen_app.py

To use the library in your own project instead, install PyTorch first and then the package. The README names Python 3.9 and PyTorch 2.0.0 for it, and says a local clone is needed to train.

pip install 'torch>=2.0'
pip install -U audiocraft
pip install -e .
sudo apt-get install ffmpeg

How to use it: Open the Space in a browser, or start your own copy with the command below. Then describe the music you want and run the model.

python demos/musicgen_app.py

Tool | What it does
MusicGen | Makes music from a text description
AudioGen | Makes sound from a text description
EnCodec | Compresses and rebuilds audio at high quality
Multi Band Diffusion | Decodes audio in a way that works with EnCodec

Example: You type a short text describing the music you want and run the demo. It returns a generated audio clip that follows your description, which you can play back.

For newcomers and experts: If you are new to this, open the Space in your browser, write one sentence about the music you want and listen to what comes back. Change a few words and run it again to hear the difference.

If you know the field, the repository bundles inference and training code for MusicGen, AudioGen, EnCodec and a diffusion decoder. The README says model files are cached by Hugging Face, and the AUDIOCRAFT_CACHE_DIR variable moves the cache for the AudioCraft models.

You write a text description
MusicGen reads the prompt
The model generates audio
You play the clip

Good to know: The Space page lists cc-by-nc-4 as its license. The README says the code is MIT and the model weights are CC-BY-NC 4.0.

• License: cc-by-nc-4 on the Space, with MIT code and CC-BY-NC 4.0 model weights
• Cost to run: none named
• Data it reaches: model files downloaded from Hugging Face and kept in a cache folder
• Not for: uses outside the CC-BY-NC 4.0 terms of the model weights

On Hugging Face, 13 November 2026

• Likes: 5,116
• Built with: gradio
• License: cc-by-nc-4.0

Links

• Space: https://huggingface.co/spaces/facebook/MusicGen
• Files: https://huggingface.co/spaces/facebook/MusicGen/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.