Published: October 4, 2026
Audio Model
Eleven v4 is ElevenLabs' current text-to-speech model, generating expressive speech from a script with inline audio tags.
It reads the same bracketed delivery tags as v3.
Key Features
Speech Generation
- Audio Tags: Bracketed inline tags such as [whispering] or [excited] shape delivery directly inside the script.
Delivery Control
- Stability: Slider from expressive and variable delivery at low values to consistent, repeatable delivery at high values.
Technical Capabilities
Details |
|
|---|---|
| Character Limit | 5,000 characters per request |
| Delivery Controls | Stability 0–1 (default 0.5) |
| Output Formats | MP3, 44.1kHz, 128kbps |
Prompting Tips
- Place audio tags like [excited] or [whispering] directly before the line they should affect, the same way you would with v3.
- Lower Stability for expressive takes, or raise it for consistent, repeatable narration.