Google introduced its latest audio model, Gemini 3.5 Live Translate. This new model is designed to deliver real-time speech-to-speech translation in over 70 languages.
Unlike previous models that wait for a speaker to finish talking, this advanced system can generate translated speech continuously while preserving the speaker's intonation, pacing, and pitch. It delivers natural, fluid audio without awkward pauses.
Gemini 3.5 Live Translate is now available for developers in public preview through the Gemini Live API and Google AI Studio. It is also available to everyone through Google Translate on Android and iOS. Support for Google Meet is expected in the near future.
The model can handle multilingual inputs without requiring users to manually configure language settings. This makes it useful for live interpretation during multilingual calls, meetings, lessons, broadcasts, and other conversations.
In Google Meet, it will enable conversations across multiple language combinations within a single meeting. This feature will launch soon in private preview for Google Workspace business users and will become available more broadly later this year.
Although the generated speech sounds natural and fluid, it will be watermarked with SynthID, helping ensure that AI-generated content remains detectable.