Google announces Gemini 3.5 Live Translate for real-time voice translation

In short: Google says its new Gemini 3.5 Live Translate can translate spoken conversations in near real time across more than 70 languages.

What happened

Google announced Gemini 3.5 Live Translate, a new AI model that turns speech in one language into speech in another language. Google says it is fast enough to follow a normal conversation, only a few seconds behind the speaker.

Unlike older translation tools that sound like a generic robot voice, Google says this one tries to keep the speaker’s tone, speed, and pitch. Think of it like having an interpreter who not only translates your words, but also keeps your speaking style.

Google is rolling it out in several places. Developers can try it in a public preview through the Gemini Live API and AI Studio, which are tools for building apps. Select business customers will get it in Google Meet starting this month, ahead of a wider release.

Google also says Gemini 3.5 Live Translate is coming soon to the Google Translate app on both Android and iOS. On Android, Google is adding a “listening mode” where you can hold the phone to your ear like a call and hear the translation, even without earbuds.

To reduce misuse, Google says every audio stream made by the system will include a SynthID watermark. This is a hidden marker inside the sound that labels it as AI-generated, and Google says it cannot be removed right now.

Why it matters

If it works as promised, this could make it easier to talk with someone who speaks a different language in everyday situations, like meetings, travel, or guided tours. It also raises trust and safety questions, since lifelike translated audio can be mistaken for a real recording, which is why the built-in watermark matters.

Source: Arstechnica