AudioCraft
FreeA single-stop code base for generative audio needs, by Meta. Includes MusicGen for music and AudioGen for sounds. #opensource
About AudioCraft
AudioCraft is a single-stop code base for all generative audio needs: music, sound effects, and compression. Developed by Meta AI, it includes MusicGen for text-to-music generation and AudioGen for text-to-sound generation. The underlying models leverage the EnCodec neural audio codec to learn discrete audio tokens from raw waveforms and use a single autoregressive Language Model (LM) to generate high-quality audio. By introducing a simple token interleaving pattern, AudioCraft efficiently models audio sequences while capturing long-term dependencies. It is open source and designed to simplify research and development in generative audio.
Key Features
Pros & Cons
- Open source and free to use
- Unified framework for multiple audio generation tasks
- Leverages state-of-the-art Meta AI research (MusicGen, AudioGen, EnCodec)
- Simplified design compared to prior generative audio models
- Generates high-quality audio with long-term coherence
- Requires technical expertise to set up and run (code base, not a web application)
- May demand significant computational resources for training and inference
- Primarily intended for research and development, not as a polished end-user tool