Gemini 3.5 Live Translate translates as a stream
Google released a speech-to-speech model that starts translating while the person is still talking, across more than 70 languages, preserving intonation, pacing and pitch.
Why it matters
Simultaneous translation stopped waiting for a pause: the say-then-wait structure that made machine interpretation unusable in live conversation is gone.
The announcement is dated 9 June 2026. The model automatically detects more than 70 languages. Unlike turn-by-turn systems that wait for the speaker to finish, 3.5 Live Translate generates speech continuously, and the translated audio stays a few seconds behind throughout the session. The output preserves the speaker's intonation, pacing and pitch. The rollout started the same day: to developers in public preview through the Gemini Live API and Google AI Studio, to enterprises in private preview through Google Meet, and to everyone through the Google Translate app on Android and iOS.