Google launched Gemini 3.5 Live Translate on 9 June. It translates 70-plus languages continuously — without waiting for sentences to finish. It preserves your voice’s tone, pitch, and pace. And it is already live in Google Translate, Google Meet, and the developer API.
Google’s Gemini 3.5 Live Translate launch is the most significant step forward in real-time voice translation since Google Translate launched in 2006. Google released the model across Google Translate for iOS and Android, Google Meet for enterprise users, and the Gemini Live API for developers — simultaneously. The model processes speech continuously, translating as the speaker talks rather than waiting for a sentence to finish. It detects which language the speaker is using automatically. It preserves the speaker’s intonation, pacing, and emotional tone in the translated output. In Google Meet, language support grows from 5 to more than 70 languages — covering more than 2,000 language combinations. The barrier between a conversation in English and one in Swahili just got measurably thinner.
What’s Happening & Why It Matters
Continuous Translation — No More Awkward Pauses
Google’s Gemini 3.5 Live Translate launch changes the fundamental interaction model for real-time voice translation. Traditional translation tools processed speech in turns. The first speaker talked. The system waited. It translated the completed sentence. Then the second speaker responded. That model introduced delays of two to four seconds per exchange — enough to break conversational rhythm and make multilingual dialogue feel stilted.
Gemini 3.5 Live Translate processes speech as it flows. The model stays just a few seconds behind the speaker — translating words as they arrive rather than after the sentence completes. By contrast, that “streaming” approach requires a model that can predict sentence structure and vocabulary ahead of completion. Additionally, the model handles background noise, overlapping voices, and informal speech patterns. Google product manager Anuda Weerasinghe and senior staff engineer Tony Lu confirmed the model targets real-world environments — busy cafes, open offices, guided tours, and ride-sharing vehicles.
Voice Quality: Your Voice, in Their Language
One previous limitation of real-time translation defined every system before this one. The translated voice sounded robotic. The flat, synthesised tone immediately signalled “machine translation” to every listener. Gemini 3.5 Live Translate specifically addresses that failure. The model preserves the speaker’s tone, pitch, and pace throughout translation. Enthusiastic speech arrives enthusiastically in the target language. A measured, deliberate speaker remains measured and deliberate. Furthermore, emotional cues carry across languages — something no translation model had previously managed consistently at scale.
All AI-generated audio carries an inaudible SynthID watermark. That watermark identifies the output as AI-generated without affecting the listening experience. In practice, it is Google‘s commitment to the same content provenance standards it demanded for video and image generation. Voice is no different.
Platform Availability — and Developer Pricing
Gemini 3.5 Live Translate is available today across three surfaces. First, Google Translate on Android and iOS gets the model immediately for all users. Second, Google Meet enterprise users on Google Workspace plans receive a private preview starting June 2026, with full rollout later in the year. Third, developers access the model through the Gemini Live API and Google AI Studio in public preview. API pricing sits at $0.023 per minute — a rate that undercuts several competing services.
Partners including Grab, Agora, and LiveKit are already building on the API. Grab — the Southeast Asia ride-hailing platform that handles millions of driver-passenger calls monthly — is testing Gemini 3.5 Live Translate for real-time driver-passenger communication across its multilingual markets.
What This Means for Apple, Amazon, and the Wider Market
Google’s Gemini 3.5 Live Translate launch arrives one day after Apple unveiled Siri AI at WWDC 2026. The timing is not accidental. Apple demonstrated strong multilingual support in the new Siri — but did not announce live voice-to-voice translation as a feature. By contrast, Google now leads on both real-time translation quality and platform breadth. Amazon‘s Alexa+ launched AI podcast generation but has not announced comparable voice translation capability. Microsoft‘s Teams already offers real-time transcription and translation — but the voice quality gap between Teams’ system and Gemini 3.5 Live Translate is likely significant.
TF Summary: What’s Next
The Gemini Live API is in public developer preview immediately. Google Meet enterprise preview begins in June. Full Google Meet rollout follows later in 2026. An Android-only listening mode — where the device continuously translates overheard speech — is available now as a separate feature. Google AI Studio provides free model access for developers evaluating the API.
MY FORECAST: Google’s Gemini 3.5 Live Translate launch will define the real-time voice translation market for the next two years. The combination of streaming translation, voice preservation, and broad platform availability at $0.023 per API minute is a competitive package no rival can immediately match. By contrast, Apple‘s WWDC 2026 absence on live voice translation is the gap that will generate the most user demand for an iOS feature in the next 12 months. Expect Apple to close that gap in iOS 27.1 or at the latest WWDC 2027 — using its existing Google Gemini licensing relationship to power the feature from the same underlying model.