Tavus's new realtime video model flirts with passing the conversational video Turing test

alexisgallagher · x · 2026-10-02

kwindla argues voice agents blew past the Turing test about a year ago: in real-world deployments like customer support, users often forget they're talking to an AI even when told upfront. He says realtime conversational video is next — Tavus just released a full-duplex realtime video model that handles vision/audio input and audio+video output in realtime, and at its best passes the nascent conversational video Turing test in both benchmarks and his own testing.

Related event: Tavus launches Griffin, real-time video conversation model that fools 48% of testers(17 posts)→

Original post →

More from Multimodal

Multimodal channel →