Google introduced Gemini 3.5 Live Translate, a new generation of real-time speech translation system. The new technology supports more than 70 languages and enables significantly more natural communication between interlocutors who do not speak the same language.
Unlike traditional translators that wait for a person to finish a sentence before they start translating, Gemini 3.5 Live Translate uses continuous translation generation. The system simultaneously listens to speech and produces translated content, so the conversation flows almost without interruption, as if in real time.
Google states that the new technology does not just translate words, but tries to retain the tone of voice, speed of speech and intonation of the original interlocutor. Thanks to this, the translated voice sounds more natural than with previous machine translation systems.
Gemini 3.5 Live Translate brings more natural conversations
The new feature is gradually coming to the Google Translate app for Android and iOS devices. A particularly interesting novelty for Android users is “Listening Mode”, an operating mode that allows listening to translated speech directly through the phone’s speakers, even when the user does not have headphones connected.
At the same time, Google is expanding its translation capabilities in the business environment as well. After integration with the Google Meet service, the new technology enables more than 2,000 language combinations for simultaneous translation during video meetings. This functionality is currently available to a limited number of Google Workspace business users through a private testing program.
The company has opened access to the technology and developers through the Google AI Studio platform and the Gemini Live API. Among the first companies to test the system is the passenger transport platform Grab, which is exploring the possibilities of easier communication between drivers and passengers on international trips.
In order to prevent misuse of AI-generated content, Google has also implemented the SynthID digital watermark, which will be embedded in all audio content created using the Gemini 3.5 Live Translate system.