Google has launched Gemini 3.5 Live Translate, a new audio model designed to make real-time multilingual conversations more natural than what existing translation tools offer. The model supports over 70 languages and is rolling out today across the Google Translate app, the Gemini Live API for developers and Google Meet for enterprise users.
Unlike conventional translation systems that wait for a speaker to finish a sentence before processing and responding, Gemini 3.5 Live Translate generates translated speech continuously, staying a few seconds behind the speaker throughout the conversation. The model automatically detects the spoken language without requiring any manual configuration.
One of the major improvements is also in how it handles voice. Rather than producing a generic synthetic voice, it preserves the original speaker’s intonation, pacing and pitch, so the translated version sounds more like the person speaking than a robot reading their words. It is also built to handle noisy, real-world environments, making it practical for settings like customer support calls, classrooms, ride-sharing pickups and live events.
The rollout spans several platforms and developers can access the model today in public preview via the Gemini Live API and Google AI Studio. In addition, Enterprise customers will get access through Google Meet starting this month in private preview, with a broader rollout to follow. The Meet integration is a notable upgrade: it expands speech translation support from the previous limit of five languages to over 70 and opens up more than 2,000 language pairings within a single meeting rather than only translating to and from English.
For regular users, Gemini 3.5 Live Translate is rolling out in the Google Translate app on both Android and iOS. Any pair of headphones can be used to hear translated audio. Android users are also getting a new listening mode that streams the translated audio through the phone’s earpiece when you hold it to your ear like a regular call which will come in handy when headphones are not at hand.
All audio output from Gemini 3.5 Live Translate carries a SynthID watermark embedded in the waveform, marking it as AI-generated. Google says there is currently no way to remove it.