Listening is visible
Live Translator does not imply background listening. The live state appears while the session is running and stops when you end it.
Speech in, translation out
Live Translator listens only during an explicit session, shows speech as it develops, and translates the finalized turn. You can read the result or play it through an available system voice.
A microphone, internet connection, and compatible device voice are required. Noise, accents, network conditions, and provider availability can affect results.
Live Translator does not imply background listening. The live state appears while the session is running and stops when you end it.
Provisional words can still change. Live Translator distinguishes them from the finalized turn that is sent for translation.
Spoken output uses an on-device system voice. Availability and voice quality vary by language and device.
On mobile, audio streams through our server to Soniox for speech recognition. Google Gemini translates submitted text and recent conversation context. With your permission, we retain submitted mobile text and its translations for product improvement for a maximum of seven days. These improvement logs do not include microphone audio. See our privacy policy for the processing and retention details.
Read the full data flow and retention boundaries.
Live Translator is not an offline translator, simultaneous interpreter, call translator, meeting recorder, or guaranteed replacement for a human interpreter. The launch product is a turn-based mobile conversation tool.
The conversation view keeps the source and translation together for the next person.