Gemini Can Now Switch Between 97 Languages in the Same Voice Chat

Gemini Can Now Switch Between 97 Languages in the Same Voice Chat

Imagine starting a conversation with Gemini in English.

Then you suddenly switch to French.

A few seconds later, you ask something in Spanish. Then you throw in another language without stopping the conversation. Instead of forcing you to change a setting every time, Gemini can now follow along.

Google has launched Gemini 3.8 Live, a new real-time voice AI model that can automatically detect and transition between 97 supported languages during a single conversation. And yes, it is designed to happen while you are still talking.

Gemini Just Got Much Better at Multilingual Conversations

For years, talking to AI in multiple languages has often meant dealing with language settings, awkward pauses, or switching between different tools. Google is trying to remove much of that friction.

With Gemini 3.8 Live, the model can detect when the language of a conversation changes and adjust automatically.

So you could start a conversation in English, switch to Yoruba or another supported language, and continue without manually restarting the conversation.

The key word here is “supported.”

Google says Gemini 3.8 Live supports 97 languages, but that does not mean every possible language or dialect is covered. Availability and specific features can vary by product and implementation.

This Is Not Just About Translation

Here is where things get more interesting.

Gemini 3.8 Live is not simply translating a sentence and sending it back. It is built as a real-time audio-to-audio model, meaning the conversation can happen directly through voice.

Google says the model is designed for fast, fluid dialogue and can continue interacting while tools and API calls run in the background. That could make conversations feel much closer to talking with another person.

You can speak. Gemini responds.

You change direction. It follows.

You switch languages. It adapts.

And the conversation keeps moving.

Google Has Been Building Toward This

This is not Google's first attempt at making Gemini multilingual.

In 2024, Google expanded Gemini Live to more than 40 languages and allowed users to set up conversations using up to two languages on the same device. Google has since pushed much further.

In June 2026, the company introduced Gemini 3.5 Live Translate, which could automatically detect more than 70 languages for near-real-time speech-to-speech translation. Google said the system was designed to preserve natural intonation and reduce awkward pauses.

Now Gemini 3.8 Live pushes the number to 97 languages while focusing on natural, continuous voice interaction.

And It Gets More Interesting for Developers

Gemini 3.8 Live is not only meant for people chatting with the Gemini app.

Google has also made the model available through its Live API, giving developers a way to build real-time voice experiences into their own applications. That could open the door to AI-powered customer support, multilingual assistants, voice agents, education tools, and other applications where conversations need to happen quickly.

Instead of building a separate voice system for every language, developers could potentially use one model capable of handling multilingual conversations.

Gemini Is Also Coming to Search Live

Google is already putting the technology to work. Gemini 3.8 Live is powering Search Live, Google's voice-based search experience, where users can have more natural conversations while exploring information.

Google says users can switch languages during these conversations while also receiving links to continue exploring information on the web. That changes the idea of what “searching” can look like.

Instead of typing a question, reading an answer, and starting again, you could simply keep talking.

Why This Matters

Language has always been one of the biggest barriers to technology. The internet may be global, but digital experiences are often still built around a relatively small number of dominant languages. Better multilingual AI could make voice technology more useful for people who naturally move between languages in everyday life.

That is especially interesting in multilingual countries and communities, where switching languages during a conversation is completely normal. The bigger opportunity is not simply getting AI to speak more languages.

It is getting AI to understand that humans do not always stay in one language.

The Bigger AI Race Is Moving to Voice

AI competition is no longer just about who has the smartest chatbot.

Companies are increasingly competing over how naturally AI can see, hear, speak, reason, and take action. Gemini 3.8 Live is Google's latest move in that direction.

The goal is simple:

Make talking to AI feel less like using a computer and more like having a conversation.

And with 97 languages now supported for automatic mid-conversation switching, Google is making a pretty clear statement.

The future of voice AI may not be about choosing a language before you start talking.

It may be about AI simply keeping up with you.