YouTube debuts an artist deepfake generator

The Future. YouTube is giving users the ability to generate “original” song clips with AI and the voices of top artists like Charli XCX and John Legend. While the feature will likely open up Pandora’s Box on the eventual creation of a legion of deepfake songs (you know people are going to hack this), YouTube may be trying to give musicians a fun way to market themselves and stay atop the short-form video conversation — a move that could be a major revenue stream for musicians and help YouTube finally catch up with TikTok’s dominance.

Short soundtracks
YouTube, with the permission of artists, is giving everyone the opportunity to become Fake Drake.

• Dream Scenario allows users to type a songwriting prompt and select a participating artist, which the tool then turns into an up-to-30-second song snippet in the voice of the artist (hear a demo here).
• Nine artists have already signed up to clone their voices: Alec Benjamin, Charlie Puth, Charli XCX, Demi Lovato, John Legend, Papoose, Sia, T-Pain, and Troye Sivan.
• The feature, built on Google DeepMind’s Lyria AI music-generation system, is currently in beta with select artists and creators to collect feedback before a wider rollout.

YouTube, which has been working with UMG and WMG on AI tools and guardrails, stresses the feature will abide by its three generative AI principles: mutual collaboration, copyright protection, and trust and safety. Also, it’s introduced requirements for labeling AI-generated content and streamlined ways for artists to request takedowns of unauthorized deepfakes.

That all may have been key for selling this to the participating artists, who are probably receiving a significant kickback for being included… and will likely dominate trending topics when the tool hits the mainstream and is on every Short imaginable.

Your turn at the mic, TikTok.