Meta AI's FAIR team released MusicGen as part of the Audiocraft library.

MusicGen is a single-stage transformer model that generates high-quality music from text descriptions or melody conditioning.

Hosted on GitHub under facebookresearch/audiocraft and on Hugging Face under facebook/musicgen models.

Released in June 2023 by Meta AI.

It provides an open, controllable alternative to closed commercial music generation APIs for research and creative use.

Built on PyTorch and EnCodec audio tokenization, runnable via the Audiocraft Python package or Hugging Face transformers.

Code released under MIT license; model weights released under CC-BY-NC 4.0, restricting commercial use; released directly by Meta AI.