Real-time voice translation has been a demo-stage promise for years. You'd see it on stage, marvel at it, and then go back to subtitles. Google is now making a serious push to change that: Gemini 3.5 Live Translate is a dedicated speech-to-speech model that any developer can wire into their app today via the Gemini Live API.
One model, three surfaces
Google launched Gemini 3.5 Live Translate as a streaming speech-to-speech model that translates across 70+ languages with automatic detection. It is rolling out simultaneously across three distinct channels:
- Developers get public preview access through the Gemini Live API and Google AI Studio.
- Enterprises get private preview access in Google Meet starting in June. Google Meet leaps from 5 English-only languages to 2,000+ combinations.
- The consumer app update is live globally on Android and iOS, with no sign-up required.
Google shipped a headphone-based beta on Android in December 2025 in the United States, Mexico, and India, expanded to iOS in March 2026, and has now delivered the remaining pieces: developer access through the Gemini Live API and Google AI Studio, plus an enterprise integration in Google Meet. The developer API is the new and most significant unlock here.
Under the hood: one model, not a pipeline
The architecture is the real story. Unlike traditional voice AI stacks that chain separate speech-to-text, language model, and text-to-speech components, the Live API processes audio end-to-end in a single model -- audio in, audio out. That matters enormously for latency. Every hop in a classic pipeline adds delay and a new failure mode. Collapsing it into one model cuts both.
The model is built on Gemini 3 Pro and streams speech-to-speech translation in over 70 languages, auto-detecting the source language and generating output that preserves the speaker's intonation, pacing, and pitch. According to the Gemini 3.5 Audio model card, it accepts audio input with up to a 128K-token context window and produces audio and text output up to 64K tokens.
Don't miss what's next in AI
Join 300,000+ engineers and researchers who get the signal, not the noise.
- Full access to in-depth AI research breakdowns
- Be the first to know what's trending before it hits mainstream
- Daily curated papers, repos, and industry moves