File photo
Washington, USA: Google has announced the launch of its new Gemini 3.5 Live Translate model for real-time speech-to-speech translation, with the aim of reducing language barriers and enabling more natural and seamless live conversations between users in different languages.
The company said the new model is capable of automatically detecting more than 70 languages, while providing voice translation that preserves the speaker's intonation, pacing, and pitch.
The model is distinguished by its ability to continuously generate translations during speech, instead of relying on the traditional translation method that waits for the speaker to finish the sentence before starting the translation.
Google showcased the model's capabilities through practical examples that included real-time video dubbing, as well as translating multilingual lectures and presentations, while maintaining a smooth flow of conversation between participants.
As part of expanding the technology's reach, the company confirmed that Gemini 3.5 Live Translate will be integrated into Google Meet later this year.
The number of languages supported for live translation will also increase from the current five to over 70, with the potential to enable more than 2,000 different language combinations in one meeting.
The service is set to move beyond its previous reliance on English as an intermediary language for translation, and Google plans to update the Google Meet interface to make it easier to access the voice translation feature more quickly.
The improved version of voice translation in Google Meet will first be available to a select group of companies participating in the Google Workspace service as part of a special preview later this month, and will expand to a wider range of users later this year.
Meanwhile, Google has begun rolling out Gemini 3.5 Live Translate globally within the Google Translate app on both Android and iOS.
The update includes a new feature called 'listening mode' in Android, which allows users to listen to live subtitles through headphones or directly through the phone's speaker.
The company affirmed that all audio clips generated by the model will include an invisible digital watermark using SynthID, a technology developed by Google to help verify AI-generated content and limit its misuse in spreading misinformation.
