YouTube unveils new AI tool to clone famous singers, with their consent

Google is in the process of testing innovative generative AI features for YouTube that will enable users to create music tracks through simple hummed tunes or text prompts. The inaugural feature, Dream Track, has already been introduced to a select group of creators on the platform. This feature is designed to automatically generate brief 30-second music tracks inspired by well-known artists. The tool can mimic the style of nine different artists who have collaborated with YouTube to aid in its development. Additionally, YouTube is showcasing new tools capable of producing music tracks from a user’s hummed tune.

YouTube is currently conducting tests on its Dream Track feature with a “small group of select US creators,” allowing them to produce tracks in the style of nine artists: Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Papoose, Sia, T-Pain, and Troye Sivan. Demonstrations reveal how simple prompts such as “A ballad about how opposites attract, upbeat acoustic” can be used to create a track in the style of Charlie Puth, or “A sunny morning in Florida, R&B” can be transformed into a T-Pain song. The software is capable of generating lyrics, a backing track, and an AI-generated voice in the style of the chosen artist.

These tracks are intended for use with YouTube’s TikTok-style Shorts service, which is also where YouTube recently announced another AI feature called Dream Screen in September. This new feature is capable of generating videos and photos to be used as backgrounds.

YouTube has revealed that participants in its Music AI incubator will have the opportunity to test these tools later this year.

These advanced AI tools are powered by a music generation model called Lyria from Google’s DeepMind. In an accompanying blog post from DeepMind, the Google subsidiary mentions that tracks created using Lyria will carry a SynthID watermark, which is inaudible to the human ear and can be preserved even when modified. This ensures that it is possible to determine if a track contains AI-generated audio from Lyria, even if it is altered in various ways.