Voice clones of non-native English speakers reportedly rated more humanlike than the originals

Announcing a NeurIPS 2026 paper, a researcher says their team cloned the voices of 86 non-native English speakers using ElevenLabs, Coqui-XTTS and ChatterBox. In 4,000 human ratings, the clones sounded warmer, more authoritative and more humanlike than the real speakers. Listeners also said they would trust the clones more. The speakers came from 22 language backgrounds; an accent classifier labeled every Coqui-XTTS clone as U.S., U.K., Canadian or Australian English.