Canadian singer-songwriter Grimes went viral late last month when she invited her fans to create music using her voice, stating that should would split 50% of royalties for any successful AI-generated song. Now, the artist has unveiled an AI voice software, called Elf.Tech, to make it even easier for users to deepfake her voice for their own AI songs. Pitchfork reports:

Artists can commercially release the results in exchange for half of any master-recording royalties. Grimes announced a pair of new songs, “Music for Machines” and “I Wanna Be Software,” in tandem with the launch, though their release date has not been set. In a Twitter thread about the software, Grimes asked users to “be tasteful” but said she would only block extreme uses, such as an AI Grimes “Nazi anthem” (“unless it’s somehow in jest a la The Producers I guess”). “Baby murder songs” are also off the menu.

Through Elf.Tech, Grimes has also shared a demo of her collaborative remake of Richie Hawtin’s Plastikman track “Passage (Out).” Find it in the “Bounces” folder on the website. You can also access stems to train your own Grimes AI. The project is powered by the generative AI Triniti.