Remix.run Logo
usef- 3 hours ago

I think they're claiming it's achieved by text models, not voice models, fwiw.