Flash v2 & v2.5 - Low Latency TTS Models

ElevenLabs has introduced two new ultra-low latency text to speech models, Flash v2 and Flash v2.5.  Achieve time to speech generation speeds under 75ms for conversational voice agents, real time game characters, and other low latency applications.

Flash v2 is English only and Flash v2.5 is multilingual, supporting 32 languages.  As with the Turbo models, at the time of release the Flash models cost 1 credit for every 2 characters generated.

For API usage, the model_is for Flash v2 is eleven_flash_v2 while Flash v2.5 is eleven_flash_v2_5

Blog Announcement: elevenlabs.io/blog/meet-flash