Meta presents new AI model that creates music based on text

The American Meta Corporation has presented a new generative artificial intelligence model, MusicGen, which creates music based on text description.

MusicGen, developed by the Audiocraft team, can create 12-second instrumental performance tracks.

We present MusicGen: A simple and controllable music generation model. MusicGen can be prompted by both text and melody.
We release code (MIT) and models (CC-BY NC) for open research, reproducibility, and for the music community: https://t.co/OkYjL4xDN7 pic.twitter.com/h1l4LGzYgf

— Felix Kreuk (@FelixKreuk) June 9, 2023

According to the developers, this AI model was trained on 20,000 hours of music from over 10,000 licensed music tracks, as well as 390,000 free instrumental tracks from the Pond5 and Shutterstock media libraries.

To get a better result, the model also allows you to upload some music as an example and specify what you want to get as a result. MusicGen can be tried for free on the Hugging Face website.

Earlier, the Google company presented a neural network that creates music based on the text description. MusicLM is available in the AI Test Kitchen app for mobile devices.