Voices are with us all the time, whether in conversations, podcasts or movies. But what if we are suddenly no longer sure whether it is a person speaking on the other end?
Today, artificial intelligence can generate voices that sound deceptively real. How does this work technically and linguistically?
Phonetics professor Volker Dellwo explains how we recognize voices and why voices are not like unchanging fingerprints. Voices are complex, changeable signals full of information about identity, emotions and intentions. We talk about forensic phonetics, lie detectors and the question of how reliable voices are as evidence.
A conversation about the science behind the voice, trust in language and how much humanity there can be in synthetic speech.
Volker Dellwo is a professor at the Institute of Computational Linguistics at the University of Zurich. He researches how humans and machines recognize and use speaker-related information in speech signals.
Further links
- NCCR Evolving Languages
- VoCS MSCA
- SRF Einstein (Sendung vom 8. Mai 2025)
Schampar digital is the podcast that communicates complex research topics from the DIZH universities in an understandable way. Our aim is to provide insights into the digital transformation and make current topics from various specialist areas tangible. More information and episode overview
Hosted by Frank Richter and Jessica Reust.
Our podcast is available on all platforms, e.g.