

AudioCraft
β Text source-checked Oct 24, 2026AudioCraft is a code base for generative audio needs including music, sound effects, and compression.
Source-checked against the vendor's site Β· rebuilt from the vendor's own pagesProfile updated Link checked How we rate
In short
AudioCraft is a free, open-source PyTorch library for generating music and sound effects using MusicGen and AudioGen models with EnCodec compression.
AudioCraft is a PyTorch library for deep learning research on audio generation. AudioCraft contains inference and training code for the AI generative models AudioGen and MusicGen.
MusicGen and AudioGen consist of a single autoregressive Language Model that operates over streams of compressed discrete music representation tokens. The models leverage the EnCodec neural audio codec to learn discrete audio tokens from the raw waveform.
AudioGen is focused on text-to-sound generation and produces audio from environmental sounds. MusicGen produces diverse and long music samples from user-provided text inputs.
AudioCraft requires Python 3.9 and PyTorch 2.1.0 for installation.
The code in the repository is released under the MIT license. The model weights in the repository are released under the CC-BY-NC 4.0 license.
According to AudioCraft: AudioCraft is a PyTorch library for deep learning research on audio generation. Its pages list this strength: Includes inference and training code for AudioGen and MusicGen. They also state this limitation: Requires Python 3.9 and PyTorch 2.1.0.
β The AI Tool Bible editorial team
Pros
- β Includes inference and training code for AudioGen and MusicGen.
- β Uses a single autoregressive Language Model architecture for MusicGen and AudioGen.
- β Code is released under the MIT license.
- β Built as a PyTorch library for deep learning research.
Cons
- β οΈ Requires Python 3.9 and PyTorch 2.1.0.
Facts
- Platforms
- web (not source-checked)
- Pricing
- free - Free and open source; self-hosted (not source-checked)
- Open source
- Yes (license: MIT)
- GitHub stars
- 23,861 (checked Nov 13, 2026) (not source-checked)
- Last GitHub push
- Mar 3, 2026 (not source-checked)
- First commit
- Jun 2023 (not source-checked)
- Parent company
- Meta (not source-checked)
- Model used
- MusicGen, AudioGen, EnCodec (not source-checked)
- Package downloads (30 d)
- 9,327 Β· pypi audiocraft Β· checked Oct 24, 2026 (not source-checked)
- Earliest Wayback capture
- 2023-08-02 (earliest archived capture of the site, not a launch date)
Where this tool sits
Use cases
Frequently asked
- Is AudioCraft free to use?
- Yes, AudioCraft is free and open source. The code is released under the MIT license, while the model weights are released under the CC-BY-NC 4.0 license. It is designed for self-hosting.
- What are the system requirements for installation?
- AudioCraft requires Python 3.9 and PyTorch 2.1.0 for installation. It is a PyTorch library designed for deep learning research on audio generation, including inference and training code.
- What specific audio tasks can AudioCraft perform?
- AudioCraft supports text-to-sound generation using AudioGen for environmental sounds and text-to-music generation using MusicGen for diverse music samples. It also includes a neural audio codec called EnCodec.
- How do the models generate audio?
- MusicGen and AudioGen use an autoregressive Language Model operating over compressed discrete music representation tokens. They leverage the EnCodec neural audio codec to learn discrete audio tokens from raw waveforms.
- Is AudioCraft suitable for commercial use?
- The code is under the MIT license, but the model weights are under the CC-BY-NC 4.0 license. This non-commercial license restricts commercial use of the weights, so users should review the specific terms.