AudioCraft (MusicGen) at a glance
Checked 16 Oct 2026AudioCraft is Meta's PyTorch library with MusicGen (300M to 3.3B parameters), MAGNeT, JASCO and AudioGen. MusicGen makes instrumental music from a text description, and the melody models can follow a reference tune. It cannot sing: the model card says it does not generate realistic vocals, so it does not replace Suno's songs. Code is MIT, but all weights are CC BY-NC 4.0, so no commercial use. It needs a GPU (16 GB recommended) and pins Python 3.9 with PyTorch 2.1.
RepoQuarry. “AudioCraft (MusicGen).” Checked 16 Oct 2026. https://repoquarry.dev/projects/audiocraftHow much of Suno does it replace?
1 of 5 jobs. Instrumental clips only; no vocals, and the weights are non-commercial.
For full songs with vocals from a prompt or lyrics, covers and restyles of an uploaded song (Cover), extending or regenerating part of a song (Extend, splitting a song into vocal and instrument stems (Get Stems), see other Suno alternatives. You may also need to build an interface, an API server and hosting yourself.
Facts, with sources
Every answer links to where we found it. We read the documentation; we did not install it.
In order to use MusicGen locally you must have a GPU.github.com · MUSICGEN.md ↗ 1d ago
The code in this repository is released under the MIT licensegithub.com · audiocraft#license ↗ 1d ago
The models weights in this repository are released under the CC-BY-NC 4.0 licensegithub.com · audiocraft#license ↗ 1d ago
Code is released under MIT, model weights are released under CC-BY-NC 4.0.huggingface.co · musicgen-large ↗ 1d ago
We recommend 16GB of memory, but smaller GPUs will be able to generate short sequencesgithub.com · MUSICGEN.md ↗ 1d ago
#### I need help on Windowsgithub.com · MUSICGEN.md ↗ today
AudioCraft requires Python 3.9, PyTorch 2.1.0.github.com · audiocraft#installation ↗ 1d ago