Humanness Index™ · TTS model
by xAI
Streaming is the WebSocket variant of xAI's Grok TTS, built for real time agents that need audio flowing before the full input has been processed.
Standings as of Jun 13, 2026, 00:15 UTC
A real arena clip: a cloned source voice reading a customer support prompt at phone quality.
Streaming is the WebSocket variant of xAI's Grok TTS, built for real time agents that need audio flowing before the full input has been processed. It accepts unbounded text over a persistent connection and begins returning audio immediately, which makes it the natural fit for live voice applications. It shares the Grok Voice stack and its five expressive voices, and it ranks alongside its REST sibling at the very top of the Humanness Index™.
Sources: docs.x.ai
Streaming shares the Grok TTS stack that reached general availability in April 2026. The WebSocket endpoint accepts unbounded text over a persistent connection, unlike the 15,000 character cap on REST requests, and bills at the same flat per character rate.
Sources: x.ai
Standings as of Jun 13, 2026, 00:15 UTC
The benchmark is open source. Suggest a model, read the methodology, or ask us to put your voice in the arena.
