Eleven v3
Audio modelPrevious generationCard summarizedOur emotionally rich, expressive speech synthesis model
Source: Eleven v3 launch post and "Eleven v3 is Now Generally Available" post
Where to call it
- ElevenLabs API
eleven_v3
Good for
- Character dialogue, audiobook narration and emotional performances
- Videos, games and media tools that need expressive delivery
- Multi-speaker conversations with interruptions and tone shifts
Overview
ElevenLabs · 06TTS, front
Eleven v3
Our emotionally rich, expressive speech synthesis model
Typical price
$0.08 / 1K chars
Eleven v3
- Max length
- ~5 min (5,000 characters)
- Languages
- 70+ languages
- Modes
- text-to-speech, multi-speaker dialogue (Text to Dialogue)
- Released
- 3 Jun 2025
Eleven v3 launched on 3 Jun 2025 as an alpha research preview built for expressive speech: inline audio tags for emotion, delivery and non-verbal reactions, multi-speaker dialogue through a new Text to Dialogue API, and 70+ languages. It reached the API on 20 Aug 2025 and left alpha on 2 Feb 2026 with better handling of numbers, symbols and notation. launch post; API post; GA post
eleven_v3- Typical price
- $0.08 / 1K chars
- Max length
- ~5 min (5,000 characters)
- Languages
- 70+ languages
- Weights
- Closed
- Released
- 3 Jun 2025
- text-to-speech
- multi-speaker dialogue (Text to Dialogue)
- audio tags
- voice cloning (instant; PVC not fully optimized)
- realtime dialogue streaming (v3 Conversational)