Eleven v4 and v4 Turbo

ElevenLabs launched Eleven v4, which it calls its most emotive text-to-speech model, and Eleven v4 Turbo for real-time voice agents, with what ElevenLabs gives as a median time to first speech of about 150ms. ElevenLabs says v4 handles multi-speaker dialogue more naturally, takes inline direction tags like [laughs], supports more than 90 languages, and makes an instant voice clone from about 10 seconds of audio. For small teams that means more directable character dialogue, and Turbo is fast enough to consider for live, voiced NPCs. Both are available now in ElevenCreative, ElevenAgents and the API. Its quality rankings come from ElevenLabs citing Artificial Analysis.

Read the official post