Published
Update
Video length:
5 min
The “True or Fake” column of 19/20 info focuses, Wednesday September 13, on audio deepfakes in the music industry. This is trickery consisting of making an artist interpret lyrics that he has never sung using artificial intelligence.
On one piece of music, we can clearly distinguish the voice of the Belgian singer Angèle, sexcept that she never performed these words. His voice is generated by artificial intelligence, and he was made to sing the lyrics of a French rap song. This trick, we call it a deepfake audio, and it’s a real phenomenon currently. While some are rather laughable, others are perfectly credible.
Increasingly easier to produce
Jason Medeiros is the rapper of the duo AllttA. A few months ago, they released the title “Savages” by recreating the voice of the American superstar Jay Z. “The first thing we had to do was write a text and record it. Then we took that audio and mixed it with the artificial intelligence model“, he explains. “I think that the fear must be oriented more towards the question of copyright, on how to legally protect artists”believes Jason Medeiros. These deepfakes audio are in any case increasingly easier to produce and more and more difficult to locate.
Find all the Vrai ou Fake broadcasts by clicking here.