Google Launches Gemini 3.5 Live Translate, a Speech-to-Speech Model That Skips Text and Keeps the Speaker's Voice

Google has released Gemini 3.5 Live Translate, an audio-native model that translates spoken conversation in real time by processing raw audio directly rather than passing it through a text pipeline, aiming to cut lag and preserve the speaker's natural voice. The model became available to developers in public preview on June 9, 2026, and is being rolled out to consumers through the Google Translate app.