Gemini 3.5
Google has announced Gemini 3.5, which primarily brings improvements to live translations. For example, translations in services such as Google Translate and Meet. The pauses currently associated with this should become a thing of the past with the arrival of Gemini 3.5. The new model supports over 70 languages and can convert spoken text into natural-sounding translated speech. Not only is the content preserved, but Gemini also attempts to incorporate the speaker’s intonation, speaking pace and pitch into the translation.
According to Google, the system works differently from many existing solutions. Whereas traditional translators often wait until someone has finished speaking before the translation begins, Gemini 3.5 Live Translate generates the translation during the conversation. This creates a smoother experience with fewer interruptions. However, the translated voice does lag a few seconds behind the original speaker.
The new technology is now being rolled out to Google Translate on Android and iOS. In the app, you can have a conversation translated instantly using the ‘Live Translate’ button. In addition, the Android version is getting a new listening mode. If you don’t have earbuds or headphones to hand, you can simply hold the phone to your ear, just as you would during a phone call, to listen to the translation.
Google Meet is also getting a major update. Whereas the video calling service previously supported only five languages for speech translation, that number is now growing to over 70 languages. This means that more than 2,000 possible language combinations are supported within a single meeting. Users can activate the feature via a new button in the Meet controls.
For business users, the feature will become available this month in a limited private preview within Google Workspace. A wider rollout is planned for later this year. In addition, developers can already get started with Gemini 3.5 Live Translate via a public preview in the Gemini Live API and Google AI Studio.
Google is also adding an extra security measure. All AI-generated audio will be marked with an invisible SynthID watermark. This ensures that the audio remains recognisable as having been generated by artificial intelligence, and Google aims to help prevent deception and misinformation.
