The voices of singers like Charlie Puth, Sia, T-Pain or John Legend will be accessible. This generative artificial intelligence will first be limited to a small group of selected users.
Published
Update
Reading time :
1 minute
:quality(50)/2023/02/21/63f4deb5b84ac_043-dpa-pa-220303-99-370938-dpai.jpg)
Create your own songs with the voice of Sia, Charlie Puth or Demi Lovato – quickly and legally? This may soon be possible on YouTube: Google announced the launch of a new generative artificial intelligence capable of creating music, including the voices of renowned artists, on November 16 In a blog post.
As with ChatGPT for text or Midjourney for images, Lyria is presented as an AI capable of generating complex and coherent melodies from simple descriptions (musical style, instruments, effects, etc.) But Google explains that it can also create short songs from A to Z with the voices of real singers, with whom the company has worked in collaboration.
Enough to encourage the creation of deepfakes ? Google ensures that its AI creations will be identifiable to leave no doubt, thanks to an audio signature “inaudible” and designed for “remain discoverable even with common modifications like adding noise or MP3 compression”. The tool will also be limited to “a limited number of creators” to begin with.
This presentation comes just after the announcement of new rules regarding deepfakes on YouTube: the platform wants to force creators to report “realistic” artificial content as such, and authorize rights holders to request the deletion of deepfakes. An important issue particularly for the music industry (with which Google explains having worked “in close collaboration”), after the multiplication of “reprises” unauthorized artists who apply the voices of other singers to a song using generative AI.