ElevenLabs launches Eleven v4: script-controlled voice acting, 10s cloning, 90+ languages

xiaohu · x · 2026-09-29

ElevenLabs released Eleven v4 and v4 Turbo, voice models that "perform the script": inline tags for emotion, pauses, laughs, whispers and sound effects plus natural-language delivery notes; Instant Voice Clone from 10 seconds of audio; context stitching and speaker stability for coherent long-form and regeneration; multi-speaker dialogue with IPA/pronunciation dictionaries; 90+ languages including Cantonese; Turbo targets real-time chat at 100ms inference and 150ms median time-to-first-audio with bidirectional streaming.

Related event: ElevenLabs Launches Eleven v4, Tops TTS Leaderboards at a Premium Price(22 posts)→

Original post →

More from Multimodal

Multimodal channel →