AudioCraft
Meta's research library for AI audio generation: MusicGen, AudioGen, EnCodec.
About
AudioCraft is Meta FAIR's PyTorch library for audio generation research, bundling training and inference code for a family of state-of-the-art models: MusicGen (controllable text-to-music — also in this catalog as a model), AudioGen (text-to-sound), EnCodec (high-fidelity neural audio codec), MAGNeT (non-autoregressive generation), JASCO (chord/melody-conditioned music), and AudioSeal (audio watermarking). 23.7k stars, MIT. It's research code — pinned dependencies (torch 2.1), ffmpeg required, expect rough edges — but it's the reference implementation the field builds on.
What it can do
- **MusicGen**: controllable text-to-music (small → large sizes)
- **AudioGen**: text-to-sound-effects
- **EnCodec**: neural audio compression codec
- **MAGNeT/JASCO**: faster and more controllable music generation
- **AudioSeal**: watermarking for AI-generated audio
- **Training code**: full recipes to train your own
Setup
```bash
pip install -U audiocraft # needs torch 2.1.0, ffmpeg
```
```python
from audiocraft.models import MusicGen
model = MusicGen.get_pretrained('facebook/musicgen-small')
model.set_generation_params(duration=8)
wav = model.generate(["A cheerful ukulele tune"])
```
GPU strongly recommended. Docs in the repo's docs/ folder.
Prerequisites
Python 3.9; PyTorch 2.1.0; ffmpeg. NVIDIA GPU with 8GB+ VRAM recommended.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: 23,689 GitHub stars (log-scaled vs catalog peers); blended with ~9,327 monthly package downloads; last push 223d ago; 0 releases/12mo; 401 open issues. Manual review: Meta's audio-gen research library (MusicGen, AudioGen, EnCodec, AudioSeal); 23.7k stars; research-grade (pinned torch 2.1, rough edges); pushed Mar 2026. License: MIT.
★ 23,689 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.6/10
- Maintenance · recency, releases, issue load
- 4.2/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · 23,689 GitHub stars (log-scaled vs catalog peers)
- · blended with ~9,327 monthly package downloads
- · last push 223d ago; 0 releases/12mo; 401 open issues
- · manual review: Meta's audio-gen research library (MusicGen, AudioGen, EnCodec, AudioSeal); 23.7k stars; research-grade (pinned torch 2.1, rough edges); pushed Mar 2026
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.