What is MusicGen?

Introducing MusicGen: a powerful single Language Model (LM) redefining the boundaries of conditional music generation, with the ability to create high-quality music by taking cues from text descriptions or melodies. Extensive studies have confirmed the superior performance of MusicGen compared to existing approaches. To witness the astonishing capabilities of MusicGen, you can explore a collection of impressive music samples and access the code at github.com/facebookresearch/audiocraft. Prepare to be amazed by the future of music composition.

In a Nutshell

• Single Language Model (LM) for conditional music generation
• Operates with compressed music tokens, eliminating the need for multiple models
• Generates high-quality samples while guided by text or melodies
• Extensive evaluation shows MusicGen outperforms baseline models
• Studies highlight the significance of each component in MusicGen

Demo

Click below to try a demo of MusicGen:

Authors

MusicGen was built by Meta, involving Jade Copet, Felix Kreuk, Itai Gat, Tal Remez, David Kant, Gabriel Synnaeve, Yossi Adi, and Alexandre Défossez.

Learn More

Cick below to access the research paper behind the AI model MusicGen: