Realtime Translation: Realtime Translation turns speech in one language into speech in another while the…
By Steven Van ·
Enterprise workspaces can stream speech through a WebSocket API and get translated speech back in a chosen voice, with optional Zero Retention Mode.
Realtime Translation, a new ElevenLabs API, turns speech in one language into speech in another while the speaker is still talking. It is a WebSocket API and is available now for enterprise workspaces, as described in the vendor's documentation.
How the connection works
You open a connection with a target_language and a voice_id, then stream audio in. Translated speech comes back in that voice. The speaker's language is detected automatically, so source_language is optional and only needs setting if you know it.
Transcripts and speed settings
- The server also sends live transcripts of the original speech and of the translated text, which can be used for captions.
- eagerness trades speed against quality. high starts translating soonest, and low waits for more of the sentence.
- The translated voice uses Eleven v4 Turbo by default. tts_model_id switches it to eleven_v3 or eleven_flash_v2_5.
Sensitive conversations
Setting enable_zrm=true turns on Zero Retention Mode for sensitive conversations. ElevenLabs describes that mode as an enterprise option that restricts logging of request inputs and outputs.
