Google has turned on voice translation that keeps pace with a conversation rather than pausing for it. On June 9 the company released Gemini 3.5 Live Translate, a speech-to-speech model that listens in one language and speaks back in another across more than 70 languages, trailing the speaker by a few seconds rather than waiting for a full stop. It is live now in the Google Translate app on Android and iOS, worldwide and free of a sign-up. For India one of the markets that tested an earlier version it arrives in a country that runs on dozens of languages at once.Live now in the Translate appThe rollout splits by audience. For everyday users the feature sits inside the Google Translate app rather than a separate product: open the app, tap Live translate in the corner, pair earphones, and speak. Google shipped it to Android and iOS together, a step on from the earlier beta that stayed on Android. Developers reach the same model through the Gemini Live API and Google AI Studio in public preview, and Google Meet gains it in a private preview for select Workspace business customers.SurfaceWho it's forAccessStatusGoogle Translate app (Android, iOS)EveryoneOpen the app, tap Live translateLive globally, no sign-upGemini Live API + AI StudioDevelopersPublic preview, ~$0.023/minPreviewGoogle MeetWorkspace business customersPrivate previewSelect customers; wider rollout H2 2026The end of turn-by-turnThe headline upgrade sits in the timing. Older voice translation worked in three steps- transcribe, translate, then speak which stacked up 10 to 20 seconds of delay and returned a flat synthetic voice. Gemini 3.5 Live Translate streams instead, processing speech as it flows and staying a few seconds behind, and it carries over the speaker's intonation, tempo and pitch so the translated voice resembles the original. Google says the model holds up against background noise and overlapping voices, the setting where older tools came apart. Jesse Hall, a developer advocate at LiveKit, called the result "effortless" multilingual voice after building a demo where everyone spoke their own language and followed each other live.Five languages become seventyThe reach is the other half. Until now, live speech translation in Google Meet handled five language pairs Spanish, French, German, Portuguese and Italian each running only to and from English. The new model lifts that past 70 languages and, in Meet, beyond 2,000 language combinations, and it drops the English-only limit so two non-English speakers can talk through it directly. Hindi sits among the supported set, alongside Arabic, Mandarin, Japanese and Korean.Two cents a minute, Meet nextBeyond the app, Google is courting builders. The Gemini Live API runs at about two cents a minute — $0.023 — which undercuts several rival services, and partners including Grab, Agora and LiveKit are already building on it. Enterprises get a slower path: Google Meet's 70-language expansion entered a private preview this month for select Workspace customers, with a wider business rollout planned for the second half of 2026. Web users see a new speech-translation control in the meeting bar.Where it still stumblesThe caveats are real, and Google has flagged the status itself. This is a preview release, so the company has yet to commit to guaranteed uptime, settled pricing or general availability. Independent benchmarks across all 70-plus languages stay absent for now, so quality claims rest on Google's own testing. Accents, fast speech and idioms still test any translation engine, and a long exchange can lose the thread of earlier context. Sessions are also ephemeral once a conversation ends the transcript disappears which guards privacy yet rules out a written record.India tested it firstIndia holds a particular place in this story. When Google ran the earlier live-translation beta six months ago, it limited the test to three markets the United States, Mexico and India putting the country among the first to use it. The contextual upgrades Google added in February, which match tone and offer idiom alternatives, still run only in the US and India. For a place where a single conversation can switch between Hindi, English and a regional language mid-sentence, a tool that translates spoken speech across 70-plus tongues without a pause speaks to daily life at the counter, in a clinic, across a sales call.FAQsWhat is Gemini 3.5 Live Translate?It is Google's AI audio model for real-time, speech-to-speech translation. It listens in one language and speaks back in another, staying a few seconds behind the speaker rather than waiting for each turn to finish.How many languages does it support?More than 70 languages, and in Google Meet over 2,000 language combinations — up from the five English-only pairs Meet supported earlier.Where can I use it right now?In the Google Translate app on Android and iOS, live worldwide with no sign-up. Developers get it through the Gemini Live API and AI Studio in preview, and Google Meet has it in private preview for select business customers.How is it different from the old voice mode?Earlier tools translated in turns, with a 10-to-20-second lag and a robotic voice. This model streams continuously and keeps the speaker's tone, tempo and pitch.Does it support Hindi and Indian languages?Hindi is among the supported languages. India was also one of the first markets in the earlier beta, and Google's contextual tone-and-idiom features run in the US and India.Is it free?The Google Translate app feature is free with no sign-up. The developer API costs about $0.023 a minute, and Google Meet pricing for the expansion stays unannounced.Is it on iPhone?Yes. Unlike the earlier Android-only beta, this release went to Android and iOS at the same time.end of article