Glossary
What is ElevenLabs?
In short
ElevenLabs is a service for speech synthesis and recognition: it reads text aloud in realistic voices in dozens of languages and turns audio into text.
Models and products · ElevenLabs
MMOOZZGG.com editorsChecked: 1 min read
For a child
Imagine a talking book: you press a page, and a kind voice reads the story, as if your grandmother were sitting next to you. A computer can read any text like that too. But a clever voice can also be made to sound like someone you know. So if someone calls in your mom’s voice and asks for a secret or for money, hang up first and call your mom back yourself.
In plain words
It is a service that turns text into speech and speech into text. You paste a text, choose a voice and a language, and in a few seconds you get audio with pauses and intonation. There is a reverse function too: the Scribe models transcribe recordings and mark which speaker is talking. The service can clone a voice (voice cloning), but the company’s rules allow it only with the explicit and informed consent of the voice owner. An example: a publisher releases an audio version of a textbook in several languages without a studio recording.
For business
Companies use such models to voice courses and videos, for voice assistants and to transcribe calls. The service is closed for Russia: ElevenLabs’ help center puts Russia on the list of countries from which the company must restrict access (checked 16 October 2026). There are no open versions of the models for your own servers. A separate risk concerns the voice itself: the usage policy forbids cloning or synthesizing the voice of a living person without their explicit consent, and voice recordings of people under 18 may not be used for a voice model. A voice agent for calls is built on technology the company can use legally, and speech quality is checked in a pilot.
For a specialist
A platform for speech synthesis (text-to-speech, TTS), speech recognition (speech-to-text) and music generation. As of 16 October 2026, the documentation lists the models Eleven v3 (more than 70 languages, up to 5,000 characters per request), Eleven Multilingual v2 (29 languages) and Eleven Flash v2.5 (32 languages, latency about 75 ms not counting the network and the app); all three support Russian. Recognition is done by Scribe v2 and Scribe v2 Realtime: more than 90 languages and word-level timestamps, and Scribe v2 also separates speakers (speaker diarization). Voice cloning comes in two kinds. Instant Voice Cloning uses a sample as a hint during generation and does not change the model, while Professional Voice Cloning fine-tunes the model on recordings. Both pass a voice check (voice captcha): it confirms that the voice owner is present and takes part, but it does not prove rights to the recording, so responsibility for consent lies with the user.
Example
An online school voices a workplace safety course with the voice of a narrator who signed consent to the creation of his voice model. A methodologist listens to each lesson before it is published.
Related terms
Questions and answers
Sources
- ElevenLabs, Do you restrict access to the service and platform for any specific countries?, checked 16 October 2026elevenlabs.io
- ElevenLabs, Prohibited Use Policy, updated 9 October 2026elevenlabs.io
- ElevenLabs, Models, documentation, checked 16 October 2026elevenlabs.io
- ElevenLabs, Voice cloning: how it works, checked 16 October 2026elevenlabs.io
Where companies use it
Not sure whether your company needs this?
AI solutions map