Meta releases new AI tool that creates music from prompts

Meta says that MusicGen was trained using an in-house library of licensed music, while AudioGen was trained on sounds available on the public domain. (Representational Image)

Meta has released a new AI tool called AudioCraft that can create "high-quality audio and music" from text prompts.

If that sounds familiar, it's because Meta had previously released a demo for MusicGen, one of the components of AudioCraft, that could generate short audio clips based on text prompts.

Also read | Meta is reportedly working on AI chatbot based on Abraham Lincoln

Meta says that it will open-source all three components of AudioCraft - MusicGen, AudioGen, and EnCodec - with a goal to, "help advance the field of AI-generated audio".

MusicGen, as the name implies, is a model that is specifically developed to create music. All a user needs to do is simply prompt it; for example - "80's rock mixed with jazz", and it will do the rest.

AudioGen is meant for creating sound effects based on prompts, and the third tool, EnCodec is a specialised decoding engine that helps with the generation of high-quality music with "fewer artifacts".

Meta says that MusicGen was trained using an in-house library of licensed music, while AudioGen was trained on sounds available on the public domain. The company is also releasing the code for all its pre-trained models.

Also read | Instagram working on labels for 'content generated by Meta AI'

"The AudioCraft family of models are capable of producing high-quality audio with long-term consistency, and they’re easy to use," wrote Meta in a blog post.

"With AudioCraft, we simplify the overall design of generative models for audio compared to prior work in the field — giving people the full recipe to play with the existing models that Meta has been developing over the past several years while also empowering them to push the limits and develop their own models".