Text-to-speech models from ElevenLabs, Hume AI, and Descript are all pushing the limits of AI-generated voice technology.

AI text-to-speech programs could “unlearn” how to imitate certain people New research shows models can be directly edited to hide selected voices, even when users specifically ask for them.

Earlier this year, Palo Alto-based AI startup Zyphra unveiled a pair of open text-to-speech models claimed to require just a few seconds of sample audio.