Fake speech: Whence it came, where it's at, and whither it's going

Fake speech has long existed in the form of vocal impersonation and accent mimicry, but in recent years audio deepfake technologies have rapidly advanced to the point where voice clones can be practically indistinguishable from bona fide human speech. This talk gives an overview of the trajectory of fake speech in forensic speech science, including how casework is expected to shift from the predominant Auditory Phonetic and Acoustic (AuPhA) method to a complementary technology-assisted (automatic) model of investigation. Futurological implications for generative voice technologies and countermeasures are also considered, with particular relevance to Singapore’s ‘Smart Nation 2.0’ framework.