Beyond acoustics: Self-relevance as a key to voice naturalness (L)

Author:
Publisher website
Google Scholar
Abstract
Synthetic voices can now achieve remarkable acoustic accuracy, yet often fail to sound “natural,” especially when designed to reproduce one's own voice. Existing frameworks define naturalness along two dimensions: deviation from acoustic norms and human-likeness. Yet these dimensions overlook the self-voice, which can feel natural or unnatural for reasons beyond the signal itself. Here, self-relevance is proposed as a complementary dimension, capturing the subjective alignment between a voice and the listener's self-representation. Evidence shows that self-relevance modulates perceived naturalness independently of acoustic match. A full understanding of voice naturalness, therefore, requires integrating physical speech properties with the listener's self-representational framework.
Funding Information
  • Fundação para a Ciência e a Tecnologia (2023.00041.RESTART)
  • Fundação Bial (BIAL 146/2020)

This publication has 23 references indexed in Scilit: