Google’s Gemini 3.5 Live Translate: Real-Time Speech Translation Is Now a Solved Problem!

TL;DR:

• Google launched Gemini 3.5 Live Translate

• Translates live speech across 70+ languages in near real-time

• Keeps your voice’s tone and pacing intact in the translation

• Live now on Google Translate

• Coming soon to Google Meet

• Open to developers today

Google has been translating words for two decades. Over a trillion words a month now, across its products. That’s not a small number. But for most of that history, the experience of live voice translation has been clunky. Wait for the speaker to finish, process, output. The conversation never quite flows.

Gemini 3.5 Live Translate changes that. It’s Google’s latest audio model, built specifically for live speech-to-speech translation that doesn’t break the rhythm of a real conversation.

So what actually makes it different?

The core technical bet here is continuous generation. Unlike turn-by-turn systems that wait for a speaker to finish before responding, 3.5 Live Translate generates translated speech on the go, balancing the trade-off between waiting for enough context to improve accuracy and translating immediately to stay in sync with the speaker. The result is that it stays just a few seconds behind the live speaker, not a full sentence behind. That gap matters more than you’d think in a real conversation.

It also preserves the feel of how someone speaks. The model produces natural-sounding translated speech that carries the speaker’s intonation, pacing, and pitch. If the person talking is animated and fast, the translation sounds animated and fast. A lot of the meaning in human communication lives in how something is said, not just what’s said. This model actually accounts for that.

It also automatically detects over 70 languages without you needing to configure anything manually.

Where you can use it

The rollout started June 9 across three surfaces. Developers can access it in public preview through the Gemini Live API and Google AI Studio. Enterprise customers will get it in private preview on Google Meet starting this month. And everyday users can find it in the Google Translate app on both Android and iOS.

The Google Meet integration is the one that’ll matter most for business users. The update expands speech translation from the previous five-language limit to 70+ languages and unlocks over 2,000 language combinations in a single meeting. Previously, it could only translate to and from English. Private preview starts this month for select Google Workspace customers, with a wider rollout later this year.

On the consumer side, the Google Translate app update adds something genuinely practical. When using the live translate feature, connect any pair of headphones and the model mirrors the speaker’s tone across 70+ languages. For Android users, there’s also a new listening mode worth paying attention to. It lets you hear translations directly through your phone’s earpiece. Hold the phone to your ear like a regular call and the translated audio streams straight to you, without needing headphones. Useful if you’re in a situation where you don’t want others to hear, or you simply don’t have your earbuds on you.

For developers

If you’re building on top of this, the Live API supports multilingual inputs without manual configuration and is designed to handle noisy, unpredictable environments. Platforms like Agora, Fishjam, LiveKit, Pipecat, and Vision Agents have already integrated it to help developers deploy voice translation apps without managing complex real-time media streaming infrastructure themselves.

Grab, the Southeast Asian super app, is testing it to enable multilingual communication between drivers and riders at pickups. A user base that makes over 10 million voice calls per month through the platform. That’s a solid real-world stress test.

The safety layer

All audio generated by the model is watermarked with SynthID, an imperceptible marker woven directly into the audio output to keep AI-generated content detectable and help prevent misinformation. Given how convincing voice AI is getting, that’s not a throwaway footnote.

Live translation at this quality has obvious implications beyond just meetings and travel. Think live broadcasts, courtrooms, classrooms, customer support. The 70-language mark and the shift away from English-only pairs is what really opens it up globally, including for multilingual African markets where code-switching between local languages and English is the norm, not the exception.

Gemini 3.5 Live Translate is available now on Google Translate (Android and iOS) and via the Gemini Live API.