Telling voice models apart from real humans is about to get even harder: Resemble AI just released a no-cost, open-source model called Chatterbox that can clone any voice using just five seconds of audio, and users prefer it 63.75% of the time over ElevenLabs. (You can try it here.) But don’t count out Hume’s new EVI 3, which “stammers anxiously, debates enthusiastically, and whispers intimately,” making it sound convincingly human. The model is also much better at picking up on the emotions in your voice, and it’ll switch up how it responds depending on your mood.
You are viewing a single comment's thread from: