What is Eleven Realtime Translation?

Eleven Realtime Translation converts incoming speech into translated audio while the speaker is talking. It offers output in more than 90 languages, with translated text and source transcripts alongside the audio.

The product is available in alpha for ElevenAPI enterprise customers. It serves developers integrating translation into contact-center calls, telecommunications services and live-event listening experiences.

Key features

Continuous speech translation

Stream source audio over WebSocket and receive translated speech in a configured target language, without waiting for a complete recording.

Translation timing control

Choose low, medium or high eagerness. Higher eagerness starts sooner; lower eagerness waits for more sentence context. Medium is the default.

Consistent output voice

Select a library voice, Instant Voice Clone or Professional Voice Clone for translated speech. The selected voice remains consistent throughout the conversation.

Text alongside audio

Receive translated text plus interim and final source transcripts for captioning or transcription interfaces.

Automatic source-language detection

Specify the source language or omit it to enable automatic detection. Each session requires a target language and output voice.

Who is it for?

Contact-center translation

Support teams could add translated audio between callers and agents who speak different languages.

Multilingual event listening

Event platforms could offer attendees translated audio feeds in their preferred languages.

Calling-service integration

Telecommunications developers could incorporate live translation into calling experiences.

How to use Eleven Realtime Translation

Contact ElevenLabs sales to request enterprise alpha access.
Choose the target language and translated output voice.
Configure the translation WebSocket session, including source-language detection and translation eagerness.
Stream incoming audio and handle translated audio, text and transcript events in your application.

Eleven Realtime Translation pricing

Contact for pricing

Contact ElevenLabs sales for enterprise alpha access and translation pricing.

Check current pricing ↗

What to know before you use it

Access is limited to enterprise workspaces during the alpha release.
Two-way conversations require application or network configuration for both directions, such as a dedicated endpoint instance for each caller.
Translated output uses the selected voice; it does not exactly reproduce each original speaker's voice across languages.
Zero Retention Mode must be enabled per request; it is disabled by default.

Frequently asked questions

Can it provide captions as well as translated speech?

Yes. Sessions provide translated text and source transcripts alongside translated audio.

Which speech model generates the translation audio?

Eleven v4 Turbo is the default. The API also permits Eleven Flash v2.5 and Eleven v3.

Can I use it for a bilingual phone conversation?

Yes, with integration work to configure translation in each direction. One session targets one language.