Model Latency & Stability Improved

Stability Improvements

Audio stability has been improved across all ElevenLabs models, most noticeably with the Turbo v2 and Turbo v2.5 models when using Web Sockets, the text-to-speech API endpoints with request stitching, Projects, Conversational AI, and the ElevenLabs Reader App.

Latency Improvements

Time to first byte latency lowered by around 20-30ms for all models.