Eleven v3
Eleven v3 introduced inline performance directions and contextual dialogue in more than 70 languages. It first appeared as an alpha preview in June 2025; the current service is generally available.
Performance tags
Tags can direct laughter, whispering, emotional tone and other vocal behavior. The release distinguished this performance-focused model from the faster Turbo and Flash models, and warned that reliable results could require more prompt refinement.ElevenLabs: first Eleven v3 preview
Practical guidance. Keep direction text separate from the spoken script in your authoring process. Listen for unintended delivery changes when the same voice moves between narration and dialogue.ElevenLabs: first Eleven v3 preview
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API, Provider product |
| Model identifiers | eleven_v3 |
| Input limit | 5,000 characters per requestElevenLabs API introduction and model specifications |
Related models
- Eleven v4
ElevenLabs · Eleven speech
Speech model for expressive narration, consistent character voices and multi-speaker dialogue.
Launch date API · Provider product - Eleven v3 Conversational
ElevenLabs · Eleven speech
Realtime v3 speech model for directed delivery in voice agents.
Launch date API · Provider product - Voice Design v3
ElevenLabs · Eleven Voice Design
Prompted voice-design model that creates and remixes synthetic voice profiles.
Launch date API · Provider product