Will.i.am Funds AI Powered Music App That Generates Songs From Texts

Former Google DeepMind researchers have launched a generative AI-powered music creation app that allows users to create songs from text prompts.

The startup is split across London and New York and has musicians Will.i.am and Tay Keith music distributor UnitedMasters among its other backers. The company also secured backing from capital firm Andreessen Horowitz.

The AI-powered Udio app understands concepts like genre and styles and can even create vocals in different languages.

The app allows users to input a natural language prompt like, “A country song about trucking.”

In a few minutes, the app generates a 30-second sample based on the desired input. Users can then extend the short samples, further refining the output by adding new sections to the song and even adding custom lyrics.

The app refuses prompts referencing copyrighted tracks. Udio also won’t generate songs if users paste in lyrics from copyrighted songs.

Udio is currently free to use as it is in beta. Current users can generate up to 1,200 songs a month. To access Udio, a Google, Discord or Twitter account is required.

Udio said it is building AI tools to “enable the next generation of music creators.”

“We believe AI has the potential to expand musical horizons and enable anyone to create extraordinary music,” the startup’s website reads.

Related:Spotify Launches AI Tool for Creating Personalized Playlists

The initial iteration of the app is powered by the Udio v1 model. The team behind it acknowledged its flaws, describing it as “capable, but not perfect.”

“We're iterating quickly, and working on longer samples, improved sound quality, supporting more languages and next-generation controllability,” according to a post on X. Upon launch, Udio was inundated with requests to generate songs. A post on the startup’s Discord said their systems were “under huge load” which led to longer generation times.

Wharton associate professor Ethan Mollick was among those given early access to Udio. He described the app’s generation abilities as “impressive.”

Udio has not disclosed the underlying training data used to power its v1 model. It would have to be licensed tracks as record labels have been aggressive in actions against AI-generated music. Among them, Universal Music Group forced Spotify to take down thousands of AI-generated tracks last summer.

Udio joins other music-related generative AI offerings, including YouTube’s Lyria model, which allows users to generate songs for videos from text prompts.