The aim of this talk is to present the challenges involved in researching voice deepfakes. Specifically, I will address the importance of studying deepfakes in the field of forensic phonetics. I will present some examples of deepfakes (and cheapfakes), considering both the artistic applications and consented uses of voice cloning, as well as the dark side of deepfakes: voice theft for commercial or criminal purposes and other non-consensual uses of cloning.
My research team aims to investigate what makes a voice truly human, with the goal of using this knowledge to design or improve current deepfake detection systems. I will present some of our latest research on deepfakes, both in the fields of perceptual phonetics and acoustic phonetics.