Overview
- Google announced the release on Tuesday, June 9, 2026, and said public previews are available to developers through the Gemini Live API and Google AI Studio.
- The model translates continuously instead of waiting for pauses, producing natural-sounding speech that stays only a few seconds behind the original speaker.
- Google is shipping the feature to consumers via the Google Translate app on Android and iOS with a new Android "listening mode" and will begin private previews in Google Meet for select Workspace customers this month with a broader rollout later in the year.
- Developer platforms such as Agora, LiveKit, Fishjam, Pipecat and Vision Agents are integrating the API to handle real-time media streaming, and partners including Grab are testing the system for live driver-passenger and meeting use cases.
- All generated audio carries an imperceptible SynthID watermark to help detect AI-created speech and curb misinformation, and the launch builds on Google’s long history of large-scale translation work that already processes trillions of translated words monthly.