Agent update
Text to speech (Eleven v4)
Eleven v4, launched on 28 Sep 2026, is ElevenLabs' more expressive text-to-speech model, with Eleven v4 Turbo as a faster version for voice agents, at a median time to first speech of about 150 milliseconds.
How to use Eleven v4 for small business
A clinic could give its phone agent a calmer, clearer tone, and use inline directions such as a gentle delivery when a patient is distressed, without changing the underlying voice. A marketing coordinator could narrate a product video or localise the same brand voice into another of the 90-plus supported languages, with the voice keeping its identity and picking up a more native accent. A training lead could turn a procedure document into an audio guide, with more reliable request stitching for longer sections in ElevenLabs Studio.
Both models are available now in ElevenAgents, ElevenCreative and the ElevenAPI. You can start from a free account. Instant Voice Clones can be made from as little as 10 seconds of audio, and Professional Voice Clones are also supported. The launch post does not list plan names or prices, and it does not mention country restrictions. Turbo is the one aimed at live agents, where earlier high-quality models were often too slow.
Most useful for: Podcast Editor – Add intros, outros, music and ad reads, Instructional Designer – Write the course content, the outline and the assessments against the agreed objectives, Content Creator – Write scripts, hooks and captions
Frequently asked questions
What's the difference between Eleven v4 and Eleven v4 Turbo?
Eleven v4 is the more expressive text-to-speech model. Eleven v4 Turbo is the low-latency version for voice agents, with a median time to first speech of about 150 milliseconds. Both are available now in ElevenAgents, ElevenCreative and the API, and you can start from a free account. The launch post does not publish plan names or prices.
Can I clone a staff member's voice from a short sample?
Instant Voice Clones can be made from as little as 10 seconds of audio, and Eleven v4 also supports Professional Voice Clones for higher-fidelity use. You should only clone a voice you have the right to use, and check ElevenLabs' terms. The post does not set out consent rules by country.
Can the same voice speak other languages without losing its identity?
Yes. Both models support more than 90 languages. A voice recorded in one language can speak the others, keeping its identity and picking up a more native accent, which the post says is stronger than before. That is aimed at dubbing and localisation, so a brand voice can be used in any supported language.