Meta open source AI language model MusicGen, which converts text and melodies into complete music

CTOnews.com, June 12 / PRNewswire-Asianet /-- Meta recently opened up its AI language model MusicGen on Github, which is based on Google's Transformer model launched in 2017. As the name of the model suggests, MusicGen is mainly used for music generation, which converts text and existing melodies into complete music.

"We used 20000 hours of licensed music to train the model and used Meta's EnCodec encoder to decompose audio data into smaller units for parallel processing, making MusicGen more efficient and faster than the same type of AI model," the R & D team said.

In addition, MusicGen also supports a combination of text and melody input. For example, you can propose to generate "a light repertoire" and ask "to combine it with Beethoven's Ode to Joy".

The R & D team also tested the actual performance of MusicGen. The results show that compared with Google's MusicLM and other music models such as Riffusion, Mousai and Noise2Music, MusicGen performs better in testing the matching of music and text cues and the credibility of composition, and is slightly higher than Google's MusicLM in general.

Meta has allowed commercial use of the model and has released a web application for demonstration on Huggingface.

CTOnews.com attach model address: click here to go

Demo application address: click here to go