Vocal analogues: A Cultural History of Technological Voice Imitation

Abstract

The impulse to simulate human voices with machines has a long history. In exploring concepts of human authenticity and simulation within mechanical and digital representations of vocality, this thesis traces the cultural history and logic of vocal imitations up to the deepfake. From the vocal analogues of the first documented mechanical speakers in the late 18th century, to the AI vocal recreation of deceased singers over 200 years later, the study uncovers a series of concepts within the ontology of what I call the technological voice. These include imitation (mechanical speakers), the transduction and codification of vocal features (forensic voiceprint), and simulation (AI-generated voice). While historically disconnected, the chapters are linked through notions of the ‘vocal composite’, where voices are constructed as socio-technical indices of human authenticity and identity, and codified as such within phenomena such as biometrics and voice synthesis. By engaging with theories of digital simulation, legal-forensic discourse and histories of technology to create a new historical framework in which to consider emerging issues of vocal simulation, this study utilises machine voices as an entry point to reflect on contemporary concerns over vocal data and the simulation of human identity, ultimately critiquing its logic.