Must Be Talking to an Angel explores the performative potential of a set of technological tools that process, modify, or imitate the human voice. These technologies widen the already existing gap between voice, body, and identity, by disconnecting the three components from each other. What can we learn from this new state of affairs about the body-voice-identity relationship? If our voice is not (or not only) our own, if it is largely filtered, but also copied and stolen, if it becomes possible to speak to disembodied voices, or to hear the voices of the deceased, how does all this, in turn, influence the notion of identity, so intimately linked to voice and body? Is there, beyond the dangers lurking, a potential for liberation in this disconnection? Can voices without a source, just like angels, tell us things we cannot say ourselves?

Voice synthesis technologies are numerous and constantly evolving. They include, among others, auto-transcription, voice recognition, voice signature, voice tuning, accent correction, modulation and cloning. They are ubiquitous, and we all use and train them, even despite ourselves or unknowingly. The implications of this proliferation of AI tools, which disrupt one of the most personal and intimate assets of the individual, the voice, are just as multiple. Psychological, philosophical, and political stakes are entangled in our vocal cords. Questions of truth and authenticity, including authorship, norms, and deviance, all the ambiguities and tensions that are inherent to the voice-identity duo, are accentuated once the voice finds itself disembodied, without a source. Added to this is the fact that biological and synthetic voices share the same world. How do they interact, inform each other, and transform each other? It is this coexistence that intrigues us the most.