Google recently announced that it’s now launched Gemini 3.5 Live Translate, a new audio model that’s capable of continuous speech-to-speech translation. Unlike traditional turn-based systems that require speakers to pause, Google says that the new model processes audio in real time, in addition to preserving the speaker’s original intonation, pitch, and pacing.
On paper, Gemini 3.5 Live Translate can automatically detect which language is currently being spoken (with support for more than 70 languages) without requiring manual configuration. It’s also designed to handle loud or unpredictable environments, and all audio generated by the model is embedded with a SynthID watermark for authenticity.
With that in mind, Gemini 3.5 Live Translate is rolling out across Google’s entire ecosystem, and will be available globally on the Google Translate app for iOS and Android. Google says that users wearing headphones will experience a mirrored tone translation, and Android users have the added benefit of “listening mode” which allows them to hold the phone to their ear to listen to translations a bit more discreetly.
Meanwhile, Enterprise users will see the feature integrated into Google Meet starting this month in a private preview for select Google Workspace customers.
