Google introduces Gemini 3.5 Live Translate for instant voice translation

Google is raising the real-time translation game with the launch of Gemini 3.5 Live Translate. This new artificial intelligence model promises to transform how we handle voice translations, maintaining the speaker's original tone, rhythm, and intonation. What's more: all of this comes with extra security, thanks to SynthID watermarks, which ensure the audio is generated by AI.

Google has been chasing the dream of real-time translation for years. Previously, it required a specific brand phone or earbud. Last year, the company took a step forward by bringing this technology to more users in the Translate app. Now, with Gemini 3.5 Live Translate, the promise is instant translation in more places and with less lag than ever. This model is part of the 3.5 family, which debuted at the I/O event, and we are already keeping an eye out for a Pro version expected to appear soon.

Gemini 3.5 Live Translate is a voice-to-voice model capable of automatically detecting and translating over 70 languages. It is fast enough to keep up with a normal conversation, staying just a few seconds behind the speaker. And best of all: the translated voice sounds more like yours than a generic robot. The demos are impressive, but you won't have to wait long to test it for yourself.

The new feature is spreading across various parts of the Google ecosystem. Developers can already start experimenting with a public preview on the Gemini Live API or in AI Studio. The model processes speech continuously and handles all multilingual inputs automatically, sparing developers from configuring everything manually. It also filters background noise in busy environments. Some enterprise customers will get access to the new translation model in Google Meet later this month, ahead of a broader rollout. And, of course, the 3.5 Live Translate will arrive soon on the Google Translate app for both Android and iOS.

Late last year, Google started testing Gemini-based live translation in the app using any headphones. Previously, it required Pixel Buds with an Android phone. With the pending update, you can use any headphones or even listen to the translation by holding the phone to your ear, as if you were on a call. For now, this "listening mode" only works on Android. Imagine listening to a near-real-time translation of a guided tour in Spanish directly through your phone's earpiece. Gemini 3.5 Live Translate audio streams are made to sound realistic, even if they don't exactly mimic your voice. But Google is being cautious. All generated audio will have integrated SynthID watermarks, marking them as AI-created, and for now, there is no way to remove this.