Understand
ElevenLabs announces Eleven v4 and Eleven v4 Turbo text-to-speech models
Overview
ElevenLabs has released Eleven v4, its most expressive text-to-speech model built on a new architecture designed to interpret tone, pacing, emotion, and context. Alongside it, Eleven v4 Turbo offers a low-latency option for real-time voice agents. The update brings improved multi-speaker conversation handling, upgraded audio tag sequencing, enhanced voice cloning stability, context stitching for audiobooks, and support for over 90 languages.
Key points
- Eleven v4 is built on a new architecture that evaluates context, emotion, and pacing to produce performed-sounding speech.
- Eleven v4 Turbo provides low latency of around 100 milliseconds with full expressive range for real-time voice agents and interactive experiences.
- Audio tag sequencing is significantly more reliable, allowing multiple tags to be stacked in a single line.
- Voice cloning features greater stability, allowing speaker identity to hold consistently across long-form work and regenerations.
- Context stitching improvements allow regenerated single lines in audiobooks to seamlessly match the surrounding chapter.
- The model covers over 90 languages, featuring major gains in Japanese, Brazilian Portuguese, Mandarin, and Cantonese, along with better IPA support.