- Gemini Live is being upgraded to Gemini Live 3.5, which handles interruptions better
- The new model can also process live visuals and blend multiple languages on the fly
- It also promises more natural and responsive conversations and the ability to access background tools
Google has upgraded its Gemini Live voice mode with a new model, Gemini 3.5 Live, which it says delivers a major improvement over the existing Gemini 3.1 Live model.
The new 3.5 model will be better at handling mid-sentence interruptions, processing live visuals, blending multiple languages on the fly, and triggering background tools. Gemini Live currently lacks the ability to access and use other tools, so this looks like a big step forward in its functionality.
Gemini Live is already one of the best things about using Gemini on your phone, letting you have natural, human-like conversations with the Google AI, but it’s not perfect. In fact, Gemini Live is pretty prone to errors.
Its flaws are probably different depending on how your voice sounds, but for me, I find it occasionally thinks I’m talking Italian and switches languages, and sometimes if I interrupt it, then it gets confused and will reset itself back to its default voice, not the one I’ve chosen (the Ursa voice is my favorite).
Hopefully the new Gemini 3.5 Live will see an end to these problems, and I'll bring you a hands-on test as soon as I have access to the new model.
Smarter translation
To coincide with the release of Gemini 3.5 Live, Google is also releasing two developer models — Gemini 3.5 Live Experimental and Gemini 3.5 Transcribe, which developers can access in the Gemini API in Google AI Studio and Gemini Enterprise Agent Platform.
Some of these powerful smart transcription features are already making their way into consumer products. On Android, Gboard’s Rambler feature uses Gemini 3.5 Transcribe to turn rambling speech into well-formatted text, automatically removing filler words such as “um” and “ah.” You can then use your voice to correct misspellings, edit the text, or change its writing style.
In the Gemini app for macOS, the model combines natural voice input with an awareness of what’s on your screen. It can call on other Gemini models in the background to summarize local files, repurpose text between apps, or generate images at your cursor, all using spoken instructions.
Perhaps most interestingly, the technology is coming soon to Chrome. You’ll be able to talk to type in any text field on the web, whether you’re replying to an email, drafting a social post, or writing a prompt for Gemini.
Gemini 3.5 Live, Gemini 3.5 Live Experimental, and Gemini 3.5 Transcribe are all being announced today. It’s not clear when they will roll out to everyone, but we’d expect it to be over the next few days.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
