The company also released the code and models for open research, reproducibility, and the music community.
“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,” tweeted Felix Kreuk, Research Engineer at Meta AI research.
“MusicGen is built on top of the EnCodec audio tokenizer. Unlike prior work, MusicGen is a single-stage transformerLM which uses an efficient token interleaving patterns, hence eliminates the need for cascading several models (e.g., hierarchically or upsampling),” he added.
Moreover, the company said that MusicGen was trained on 20,000 hours of music, including 10,000 “high-quality” licenced music tracks and 3,90,000 instrument-only tracks from the ShutterStock and Pond5 stock media libraries.
However, Meta isn’t the first to offer an AI-powered music generator tool.
In May, Google released ‘MusicLM’ — a new experimental AI tool that can generate high-fidelity music in any genre given a text description.
The tool was first announced in January this year and is now available to the public.
The text-to-music AI tool is available in the AI Test Kitchen app on the web, Android or iOS.