Music startup Suno Inc. has unveiled an updated AI model v6 that no longer allows users to mention real artists when generating songs. The model, developed in collaboration with leading music market players, generates tracks faster and introduces a mechanism for paying royalties to copyright holders.
Music startup Suno Inc. has unveiled an updated AI model v6 that no longer allows users to mention real artists when generating songs. The model, developed in collaboration with leading music market players, generates tracks faster and introduces a mechanism for paying royalties to copyright holders.
As Suno’s chief product officer Jack Brody told Bloomberg, the v6 model will replace previous versions of the service, which did not provide for royalties for content creators. Now, the startup will transfer royalties to labels and publishers, who will then distribute the funds between artists and songwriters.
In addition to restrictions on artist names, the new v6 model has expanded capabilities: users can create compositions based on uploaded audio, images, and videos, as well as combine multiple genres in a single query.
Warner Music Group and independent publisher BMG Rights Management have already become partners in the project. According to Warner Music CEO Robert Kirkle, Suno’s payments per user will exceed the revenues the label receives from streaming services of the level of Spotify.
«The future of music must be built in partnership with the music industry and artists. This is a defining moment for Suno, marking the beginning of a new era,» said Jack Brodie.
The startup is also currently working on a separate tool for creating remixes and covers of songs using AI. The release is being delayed as musicians seek additional guarantees of protection and compensation for the use of their voices.
Suno has yet to sign deals with Universal Music Group and Sony Music, two of the world’s largest record labels. Both labels have previously filed lawsuits against the startup, accusing it of illegally using a catalog of copyrighted songs to train neural networks.
Meanwhile, data obtained from the Suno hack confirmed that the startup trained its models by mass-collecting millions of songs and lyrics from popular platforms, including YouTube Music, Deezer, and Genius. Since the company has always avoided revealing its training sets, this leak was a rare opportunity to learn where the developers actually got their content from.
