AudioCraft (MusicGen)
Meta's PyTorch library for audio generation, home of the MusicGen text-and-melody music model.
| Language | Python |
|---|---|
| Category | ML & generative |
| License | MIT |
| Platforms | macOS Windows Linux |
| Install | pip install -U audiocraft |
| First released | 2023 |
| Maintained | Yes |
| Links | Website · Source |
Strengths
- High-quality text/melody-conditioned music generation
- Also ships AudioGen, EnCodec, MAGNeT under one library
Limitations
- No release since v1.3.0 (May 2024); appears stalled
- Model weights are CC-BY-NC (separate from MIT code)
Best for
Text-prompted music and sound-effect generation and neural-codec research.
Minimal example
from audiocraft.models import MusicGen
model = MusicGen.get_pretrained('facebook/musicgen-small')
wav = model.generate(['upbeat happy rock'])Alternatives to AudioCraft (MusicGen)
See all AudioCraft (MusicGen) alternatives & how they compare →