Page 27 - Tvorjenje pomena in večrazsežna – večkodna pismenost

Jure Longyka Distinguishing Between Natural and Synthesised Speech
Radio Slovenia, Slovenia in Slovenian
jure.longyka@gmail.com
This presentation reports on a study measuring how well listeners can distinguish
Maja Krebl
natural from synthesised Slovenian speech, given current speech-synthesis qual-
maja.krebl@gmail.com
ity, and which factors shape that perception.
Marko Bajec Distinguishing is more successful among respondents professionally engaged
University of Ljubljana, Slovenia with electronic media and those with prior experience with synthesized speech.
marko.bajec@fri.uni-lj.si Announcers and actors outperform other respondents (Barrington et al., 2025; La-
©2026JureLongyka,MajaKrebl, van et al., 2019; Rosi et al., 2025). Distinguishing varies significantly across the four
and Marko Bajec speakers represented. The hypothesis that familiarity with the speaker improves
distinguishing was suggested in individual cases but statistical analysis did not
confirm it.
Barrington,S.,Cooper,E.A.,&Farid,H.(2025).Peoplearepoorlyequippedtodetect
AI-powered voice clones. Scientific Reports, 15, 11004.
Lavan,N.,Burton,A.,Scott,S.K.,Scott,S.K.,&McGettigan,C.(2019).Flexiblevoices:
Identity perception from variable vocal signals. Psychonomic Bulletin & Re-
view, 26(1), 90–102.
Rosi V., Soopramanien, E., & McGettigan, C. (2025). Perception and social evalua-
tion of cloned and recorded voices: Effects of familiarity and self-relevance.
Computers in Human Behavior: Artificial Humans, 4, 100143.

Meaning-Making, Multiliteracies
and Multimodality
Abstracts of the International
Symposium
Koper, 19–20 March 2026

https://doi.org/10.26493/978-961-293-565-8.24 27