Close ad

Imagine talking to a stranger, but instead of a robotic voice from your mobile phone, you hear a fluent translation directly in your earpiece. Moreover, with the intonation and voice timbre of the original speaker. Google translator has just launched a technology that will definitively break down language barriers. Machine translation at Google is celebrating its twentieth anniversary this year, and the company boasts that its products now translate over a trillion words per month. New audio Model Gemini 3.5 Live Translate aims to take the entire discipline to the next level. Forget the traditional text translation and the awkward concept of having to say a sentence, wait for it to be processed, and then play the result. Google is moving towards seamless simultaneous interpreting, the kind we know from professionals.

The new model can translate spoken language in virtually real time, in more than 70 languages. Unlike older solutions, it does not wait for the speaker to finish speaking. It translates continuously, staying behind the speaker for just a few seconds, while maintaining their intonation, pace, and pitch. The resulting translation simply sounds like the original speaker. The key innovation is continuous speech generation. The model constantly intelligently balances whether to wait for additional context (and thus improve the quality of the translation) or translate immediately and keep up with the speaker. The result is fluent speech without unnatural pauses. In addition, there is no need to manually set the language. The model detects over 70 languages ​​on its own and can easily cope with multilingual conversations or noisy environments.

In the new “listening mode”, you simply hold the phone to your ear, as if you were talking to someone on the phone. You can then hear the translation discreetly directly from the phone’s receiver, which completely changes the dynamics of personal communication in a foreign environment. No more holding your hand with the display in front of a stranger’s face. The new feature will gradually be introduced in video calls in Google Meet. Speech translation will arrive there first as part of private testing for select Google Workspace business customers, with Google promising wider deployment later this year. The model is available to developers via Gemini Live API. Among the first to test it is, for example, the Asian transport platform Grab for communication between drivers and passengers - over 10 million voice calls are made monthly through its application alone. The platform's management mainly praises the automatic language detection and extremely low latency.

Technology that can speak a foreign language in your own voice is, of course, a double-edged sword. To prevent the spread of deepfakes and disinformation, all audio generated by the model carries an inaudible digital watermark called SynthID. This allows for reliable AI content recognition in the future. With this step, Google is significantly strengthening its competitive edge. Apple recently bet on live translation in AirPodech and Samsung are massively pushing their AI interpreter directly into their phones Galaxy. However, the new product from Google works perfectly with Czech and Slovak, and the translation quality is top-notch according to our first tests. Manufacturers of single-purpose hardware translators will have an increasingly difficult time.

How to activate the function immediately?

  • Update the Google Translate app in the app store (Google Play / App Store).
  • After opening the application, tap the "Conversations" icon.
  • Select the new item "Listening" and click "Start".
  • Put the phone to your ear and start listening.chat surroundings.

According to our first editorial tests, the feature works very reliably and imitates the voices of the original speakers surprisingly faithfully. What do you think of this sci-fi novelty? Share with us in the discussion.

Today's most read

.