Éva Székely

Outside the Voice Box: Rethinking AI Voice Design Beyond Embodied Constraints

As AI-generated voices approach human-like realism, they also open new possibilities for vocal expression beyond biological constraints. This work-in-progress (co-authored by Éva Székely and André Pereira) presents a survey-based experiment investigating four dimensions of disembodied flexibility in Text-to-Speech (TTS) systems: Essence, Expand, Palette, and Ombré. These concepts explore how synthetic voices can transcend static identities by introducing dynamic transformations, interpolating between vocal characteristics, and expanding expressivity beyond natural human speech. We examine whether such flexibility can reinforce AI voices as distinctively synthetic, avoiding deceptive realism while offering users new forms of vocal agency. Through a survey evaluating these concepts across twelve common TTS applications, we gathered both quantitative ratings and qualitative insights on their perceived suitability. Our findings contribute to speculative voice design discussions, engaging posthumanist perspectives to challenge assumptions about AI voice identity and expressivity.