AudioCraft

What AudioCraft does

• MusicGen. Generates music from text prompts, producing diverse, long samples.
• AudioGen. Generates environmental sound effects from text descriptions.
• EnCodec codec. Neural audio codec converting raw audio to discrete tokens and back to waveform.
• AudioSeal watermarking. Included in the repository for watermarking generated audio.
• Training and inference code. Library includes code to run and train AudioGen and MusicGen models.

Who it’s for

• Researchers experimenting with generative audio models
• Developers who want to self-host text-to-music code
• Engineers training audio models on their own data

AudioCraft pricing

Free and open source; code under MIT license, model weights under CC-BY-NC 4.0.

As listed on November 9, 2026. Check audiocraft.metademolab.com for current prices.

Things to know

• Model weights are CC-BY-NC 4.0, which restricts commercial use
• Requires Python 3.9 and PyTorch 2.1.0; code-only library with no hosted app
• No official consumer interface; used through code or the GitHub repository

Quick answers

What is AudioCraft?

AudioCraft is a PyTorch library from Meta AI for audio generation research, covering inference and training code for text-to-music and text-to-sound models.

Is AudioCraft free?

Yes. AudioCraft is free to use. (As listed on November 9, 2026.)

Who makes AudioCraft?

AudioCraft is made by Meta AI.

Where can you use AudioCraft?

AudioCraft is available on desktop (apps, editor plugins or browser extensions) and an API for developers.

What are some alternatives to AudioCraft?

Other tools filed in the same section include AIVA, Soundraw and Mubert.

Tags

#musicgen#audiogen#encodec#open-source#pytorch#text-to-audio