Eleven v4

Eleven v4, launched September 28, 2026 alongside its speed-optimized sibling Eleven v4 Turbo, is a ground-up rebuild of ElevenLabs' speech synthesis architecture and the successor to v3. The headline shift is from reading text to performing it: the model reads the whole script for context, keeps speaker identity stable across long narration, and follows inline stage directions such as [softly] or [laughter] typed directly into the text. Language support grew from roughly 70 languages to more than 90.

v4 targets long-form and creative production such as audiobooks, dubbing, and character voiceover, while v4 Turbo is built for real-time voice agents with a reported median time to first speech of about 150 milliseconds over WebSocket. At launch it ranked first on Artificial Analysis' Voice TTS Arena leaderboard. Instant Voice Clones can be created from about 10 seconds of reference audio. The launch coincided with ElevenLabs reaching a $22 billion valuation.