Eleven v4: ElevenLabs Text to Speech Model, Turbo, Languages and Pricing

What Eleven v4 is

Eleven v4 is a text-to-speech model from ElevenLabs, launched on 28 September 2026 together with a low-latency variant, Eleven v4 Turbo. ElevenLabs says it is built on an entirely new architecture and was designed to interpret tone, pacing, emotion, character and context from the text, so the same line can be delivered gently, urgently, comically or conversationally while the speaker's identity stays the same. Both models are available in ElevenAgents, ElevenCreative and through the ElevenLabs API.

Directing delivery

Users can describe in natural language how a line should be delivered, and add inline audio tags for phrasing, emotion and sound effects, such as [laughs], [said angrily in French accent], [light rain] or [phone buzzing]. ElevenLabs says Eleven v4 follows these tags and direction prompts more closely than earlier models, and its developer documentation covers the full tag syntax and how to apply tags through the API. Support for International Phonetic Alphabet phonemes has been improved, so custom pronunciations behave more reliably.

Dialogue, voices and languages

Eleven v4 uses a new method to capture each speaker's identity, which ElevenLabs says keeps voices consistent across agent conversations, audiobooks and ads, and across generations, narration and regenerated lines. Because the model reads the context of a whole scene, speakers in a multi-speaker dialogue respond to what was just said rather than reading isolated lines. Both v4 models support more than 90 languages, and a voice recorded in one language can speak another with a native accent while keeping its identity. Instant Voice Clones can be made from 10 seconds of audio, Professional Voice Clones are supported, and chaining generations for long-form content is more reliable in ElevenLabs Studio and the Reader app.

Intended uses

ElevenLabs names audiobooks, character performances, voiceovers, dubbing and localised conversational agents as the content Eleven v4 and Eleven v4 Turbo are built for. For dubbing and localisation, it says a chosen brand voice, whether a specific actor or a house style, keeps its identity across every supported language. It also lists more natural multi-speaker dialogue for games and narration, where characters need to react to each other within a scene.

Eleven v4 Turbo

Eleven v4 Turbo applies the same technology to low-latency uses such as voice agents. ElevenLabs gives a median inference latency of about 100 milliseconds and a median time to first speech of about 150 milliseconds over WebSocket streaming, measured in September 2026 with network latency removed. ElevenLabs says it optimised Eleven v4 Turbo and its ElevenAgents platform together as one system, and gives healthcare agents that pronounce medical terms and gaming characters as example uses.

API pricing

On the ElevenLabs API, Eleven v4 text to speech costs $0.08 per 1,000 characters, the same rate ElevenLabs lists for Eleven v3. Eleven v4 Turbo costs $0.04 per 1,000 characters. Billing is by the number of characters converted, so a 10,000-character script costs $0.80 with Eleven v4 and $0.40 with Eleven v4 Turbo. ElevenLabs offers the API on subscription plans billed monthly or yearly and on pay-as-you-go.

Official Resources