According to unite.ai, ElevenLabs has released its text-to-speech models Eleven v4 and its low-latency version Eleven v4 Turbo, claiming them to be the company's most expressive models to date. Both new models are now available on the ElevenAgents, ElevenCreative, and ElevenAPI platforms and can be used with free accounts. Built on a brand-new architecture, Eleven v4 can analyze tone, rhythm, emotion, character, and context in text, and introduces a new speaker identity capture method to maintain consistent vocal tones. It also supports scene-level context for more natural conversations. Users can control the delivery using natural language and inline audio tags, while SSML tags have been discontinued in v4. Both models support over 90 languages and feature instant voice cloning, requiring only 10 seconds of audio for high-fidelity capture, with all clones requiring verification and authorization from the voice owner. Eleven v4 ranked first on the Artificial Analysis Provider Voice Arena leaderboard in September 2026, with pricing following the existing credit structure and a free allowance of 10,000 credits per month.