MusicGen is a major innovation launched by Meta. Developed by Meta’s audio research team, Audiocraft, MusicGen is an AI language model capable of generating music from text descriptions.

It was trained on 20,000 hours of music, including 10,000 high-quality tracks and 390,000 instrumental tracks.

Thanks to AI, MusicGen is able to understand users’ textual queries and generate original musical excerpts based on them. Whether by specifying the desired track type or by humming a melody, MusicGen is able to produce musical variations in response to these inputs.

Open-source: MusicGen was made public by Meta in June 2023. Its code is available as open-source on Github, thus expanding its accessibility.
Superior performance: Studies have confirmed MusicGen’s superior performance compared to existing approaches. MusicGen is positioning itself as a serious competitor to MusicLM, the AI ​​melody generation tool unveiled by Google, capable of composing a song in response to a prompt and/or a melody.

—

Do you need more details on a specific aspect?