Eleven v4 Turbo: ElevenLabs Low-Latency Text-to-Speech Model, Agents, Languages and Pricing

What Eleven v4 Turbo is

Eleven v4 Turbo is a text-to-speech model from ElevenLabs, released on 28 September 2026 alongside Eleven v4. ElevenLabs describes Eleven v4 as its most emotive text-to-speech model, built on an entirely new architecture, and Eleven v4 Turbo as the low-latency variant that brings the same technology to real-time uses such as voice agents. In the API it is called eleven_v4_turbo. ElevenLabs lists support agents, AI assistants and interactive characters as the scenarios it is built for, and recommends it for real-time conversational applications.

Latency

ElevenLabs gives Eleven v4 Turbo a median inference latency of about 100 milliseconds, which it says is shorter than the average pause between two people talking, and a median time from request to audible speech of about 150 milliseconds. It states that it measured the second figure in September 2026 with identical scripts and default settings against Cartesia Sonic 3.6, xAI TTS, Google Gemini 3.8 Flash-Lite TTS and OpenAI GPT-4o mini TTS, with network latency removed, using Eleven v4 Turbo over WebSocket streaming. In the API documentation the model is available through the Text to Dialogue websocket.

Expression and control

Like Eleven v4, the Turbo model is meant to interpret tone, pacing, emotion, character and context from the text rather than reading it flatly. Delivery can be described in natural language, and inline audio tags such as [laughs], [said angrily in French accent], [light rain] or [phone buzzing] set emotion, accent or sound effects for a phrase; ElevenLabs states that v4 follows these tags and direction prompts more accurately than earlier models. Support for International Phonetic Alphabet phonemes was improved for custom pronunciations, and a new method for capturing speaker identity keeps voices consistent across multi-speaker conversations.

Languages and voice cloning

Eleven v4 and Eleven v4 Turbo support more than 90 languages, from Afrikaans and Amharic to Cantonese and Cebuano; the language_code parameter takes ISO 639-1 codes, with ISO 639-3 codes for languages that lack one. ElevenLabs states that a voice recorded in one language can speak the others fluently with a native accent while keeping its identity, and that the voice no longer drifts back to its source accent during a long generation. Instant Voice Clones can capture a voice from 10 seconds of audio, and the v4 family adds support for Professional Voice Clones for the highest-fidelity cloning.

Availability and pricing

Both v4 models are available in ElevenAgents, ElevenLabs' conversational agents platform, in ElevenCreative and through ElevenAPI, and a free account can generate with either. ElevenLabs states that its teams optimised Eleven v4 Turbo and ElevenAgents together as one system. On the ElevenLabs API, Eleven v4 Turbo costs $0.04 per 1,000 characters, about $0.04 per minute of audio, compared with $0.08 per 1,000 characters for Eleven v4. Subscription plans include a monthly allowance of characters, from 20,000 on the smallest paid tier upward.

Official Resources