Gemini 3.5 Live Translate listens to spoken words and renders them in another language in near real time. Google announced it as a dedicated live speech-to-speech model in June 2026 and named three rollout surfaces: Google Translate, Google AI Studio / the Gemini Live API, and speech translation preview in Google Meet.
Picture a 9 a.m. standup where your engineer in São Paulo speaks Portuguese, your designer in Seoul speaks Korean, and you're following along in English. A live translator that keeps up while everyone is still talking changes that meeting from "wait for the recap" to "respond in the moment." That's the promise behind Gemini's live translation — and it's a real, useful one.
You already suspect the catch: a feature this good is usually tied to the ecosystem that ships it. This guide explains what Gemini live translation actually is, how the live loop works, which languages and platforms it reaches, whether it's free, and — honestly — where it stops being the right tool. If you live across Zoom, Microsoft Teams, and in-person conversations, you can open MirrorCaption in your browser and read along without waiting for Google to support your platform.
- What it is: Gemini 3.5 Live Translate converts speech to another language in near real time through Google Translate, Google AI Studio / the Gemini Live API, and Google Meet speech translation preview.
- How it works: a continuous loop of listen, transcribe, translate, and speak, so the output keeps pace with the conversation instead of waiting for it to end.
- The boundary: it works best inside Google's own products; it does not natively translate a Zoom or Teams call, and the live output is mostly ephemeral.
- The alternative: a browser-based tool like MirrorCaption captures any browser meeting tab or in-person audio, keeps a searchable side-by-side transcript, and can speak the translation aloud.
What is Gemini live translation?
Gemini 3.5 Live Translate is Google's real-time spoken-translation model: you speak in one language, and the system produces the meaning in another language as live text or spoken output, fast enough to hold a conversation. It is the live, voice-first cousin of typing a sentence into a translation box.
One naming note worth getting straight. Google announced Gemini 3.5 Live Translate in June 2026 as an audio model for live speech-to-speech translation, detecting 70+ languages and producing natural-sounding translated speech that preserves the speaker's intonation, pacing, and pitch. Google has said it is rolling out across Google AI Studio, Google Translate, and Google Meet. The Gemini app and Pixel Live Translate are related Google translation experiences, but they should not be described as the same documented 3.5 Live Translate rollout unless Google shows that exact capability in your account. For the exact model version and where it's live, Google's official pages are the source of truth.
The reason this matters for you is simple: where a feature lives determines where you can use it. Gemini 3.5 Live Translate is strongest when the conversation happens inside a supported Google surface. The moment the conversation moves somewhere else, the answer gets more complicated.
How does Gemini live translation work?
Under the hood, live translation is not one step — it's a fast loop that repeats many times a second. Understanding the loop explains both why it feels instant and why it sometimes stumbles.
- Capture: the microphone (or a call's audio) streams speech into the model continuously, rather than waiting for you to finish a sentence.
- Transcribe: speech is converted to text in the source language, with early guesses that get corrected as more context arrives.
- Translate: Gemini's language model maps the meaning into the target language, using surrounding context so idioms and tone survive better than a word-for-word swap.
- Render: the result appears as live text and, in voice mode, is spoken aloud in the target language.
- Refine: as the speaker continues, earlier output can be adjusted so the running translation stays coherent.
The big advantage of a streaming loop is that you read or hear the translation while someone is still speaking. The trade-off is that early words can shift as context fills in — a normal property of every real-time system, not a defect unique to Google. If you want a deeper look at why streaming output behaves this way, our explainer on how accurate AI translation is breaks down the accuracy-versus-latency trade-off.
What languages and platforms does Gemini live translate support?
Google says Gemini 3.5 Live Translate detects 70+ languages, but exact coverage depends on which surface you're using. Google Translate, Google AI Studio / the Gemini Live API, and Google Meet preview can each expose language support on different timelines. Because that list moves, the honest answer is to check Google's own help pages for the language you need rather than trust a fixed number from a blog post.
Platform coverage is the more important story. Gemini 3.5 Live Translate is designed to run inside Google's ecosystem:
- Google Translate app (Android and iOS): the consumer app surface Google named for rollout.
- Google AI Studio / Gemini Live API: the developer surface for testing and building with gemini-3.5-live-translate-preview.
- Google Meet speech translation preview: meeting translation for eligible Workspace accounts as Google rolls the feature out.
Priya, a product manager, runs a weekly sync with teammates in Bangalore and Berlin. She can test Google live translation in the Translate app and watch for Meet preview access in her Workspace account. But her actual standups run on a Zoom link her client controls — and that's where the Google-only boundary bites. The feature she wants simply isn't present in the call where she needs it most.
Is Gemini live translate free?
Partly, and it depends on the surface. The Google Translate app is the consumer route to check first. Developer access through the Gemini Live API has its own pricing, preview limits, and quota rules. Google Meet speech translation depends on eligible Workspace preview access. Google adjusts what's included fairly often, so the safest move is to confirm on Google's current pricing pages before you build a workflow around a specific capability.
The practical takeaway: "free" usually covers casual, occasional translation, while continuous, meeting-grade live translation tends to sit behind a subscription. That's a reasonable model — it's just worth knowing before you assume the feature is free for the way you actually work.
Where Gemini live translation falls short
To be fair, when you're inside Google's world, this is one of the smoothest live-translation experiences available. The shortcomings aren't about quality — they're about reach and what you get to keep.
It's tied to Google's ecosystem
Gemini live translation lives in Google apps and on Google's devices. If your meeting is on Zoom, Microsoft Teams, or Webex, Gemini isn't in the room. Plenty of teams don't get to choose the platform — the client or host does — so "it works great in Google Meet" can be irrelevant on the call that matters.
The live output is mostly ephemeral
Spoken live translation is designed to be heard in the moment. That's great for flow, but it leaves you without a clean, searchable record of who said what in both languages. If you've ever needed to re-read a translated sentence after a negotiation, you've felt this gap. (For the distinction between fleeting captions and a kept record, see our piece on the difference between live captions and transcripts.)
It's one tool for one ecosystem, not every situation
A doctor's visit, a rental-contract conversation, a sales call on a platform you don't control — these don't all happen inside Google products. A single-ecosystem feature can't follow you across every meeting tool and every in-person moment.
When a browser-based translator fits better
If your conversations stay inside supported Google surfaces, Gemini 3.5 Live Translate is a strong default — use it. The case for a browser-based tool like MirrorCaption starts the moment you cross platforms or need to keep what was said.
MirrorCaption isn't a captions-only reader. With Speak Translations, you speak in your language and it can read the translation aloud in theirs — through your laptop speaker, a paired phone, or, on the Mac client, a virtual microphone that routes the spoken translation into Zoom, Meet, or Teams. That's the same near-real-time, cross-language exchange people want from live translation, but not locked to one ecosystem.
For in-person conversation, Talk mode on your phone runs as one continuous session — not a push-to-talk button. You start it once, both sides take turns, and the transcript and translation context carry across the whole exchange. And because MirrorCaption captures the meeting tab's audio in desktop Chrome or Microsoft Edge, no bot ever joins the call; participants don't have to approve a meeting recorder. If you specifically want real-time translation for Google Meet that also follows you to other tools, that's exactly the gap it fills.
| Gemini live translation | MirrorCaption | |
|---|---|---|
| Where it works | Google Translate, Google AI Studio / Gemini Live API, and eligible Google Meet preview | Any browser meeting tab (Zoom, Teams, Meet, Webex) in desktop Chrome or Edge, plus in-person on a phone |
| What you keep | Mostly live, in-the-moment output | Side-by-side original and translation you can search and export |
| Spoken translation | Supported by the live-audio model/API and selected product surfaces | Optional Speak Translations, including a Mac virtual mic for meetings |
| Setup | Built into supported Google surfaces or custom API apps | No install and no bot — open a browser tab |
| Pricing model | Depends on Google Translate availability, Live API pricing, and Workspace preview access | 1 free hour, then €54.99/year (100h) or €99 one-time Premium (200h); Voice Packs from €2.99/5h sold separately |
On pricing specifically: MirrorCaption's €99 Premium is a one-time plan — no recurring subscription — that includes 200 hours of hosted transcription credit and all future updates, and gives Premium customers the lowest per-hour rate when they top up with Voice Packs. It is not unlimited hosted time; once the included hours run out, more hours come from Voice Packs (sold separately).
Mateo, a freelance designer in Lisbon, takes client calls on whichever platform each client prefers — Zoom this week, Teams the next. He opens MirrorCaption in a browser tab, reads the live English translation of his Spanish-speaking client side by side with the original, and exports the transcript afterward to double-check a deliverable detail. The point isn't that Gemini is bad; it's that Mateo can't ask every client to switch to Google Meet.
Curious how the live experience holds up across languages? Try it on your next call: open MirrorCaption in your browser — 1 free hour, no credit card, no install.
Frequently Asked Questions
What is Gemini 3.5 Live Translate?
It is Google's dedicated live speech-to-speech translation model, announced in June 2026, with rollout across Google Translate, Google AI Studio / the Gemini Live API, and Google Meet speech translation preview.
How does Gemini live translation work?
Gemini listens to speech, transcribes it, translates the meaning with its language model, and produces the result as on-screen text and synthesized speech. The loop runs continuously, so the translation keeps pace with the conversation rather than waiting for the speaker to finish.
What languages does Gemini live translate support?
Google says Gemini 3.5 Live Translate detects 70+ languages, but exact language support depends on the surface: Google Translate app, Gemini Live API / Google AI Studio, or Google Meet preview. Check Google's own documentation before relying on a specific pair.
Is Gemini live translate free?
Consumer availability depends on the Google Translate app and your region. Developer API use has separate pricing and preview limits, while Google Meet speech translation depends on eligible Workspace preview access.
Can Gemini translate a Zoom or Teams meeting?
Gemini's live translation lives inside Google's own products, so it does not natively translate a Zoom or Microsoft Teams call. For those platforms, a browser-based tool like MirrorCaption captures the meeting-tab audio without a bot joining the call.
Gemini live translate vs Google Translate — what's the difference?
Google Translate is the consumer app surface Google named for rollout. Gemini 3.5 Live Translate is the underlying live-audio model/API capability; the Gemini app and Pixel Live Translate are related Google translation experiences but not the same documented 3.5 rollout surface.
The bottom line
Gemini 3.5 Live Translate is a genuinely good real-time capability — fast, voice-first, and smooth where Google exposes it through Google Translate, the Live API, and eligible Google Meet preview. If that's where your conversations live, it's an easy yes. Just remember three things: the exact model version and language list move often, so check Google's pages; access and pricing depend on the surface; and the live output is built to be heard, not kept.
The real decision is about reach. If you work across Zoom, Microsoft Teams, Webex, and face-to-face conversations — and you want a searchable, exportable record plus optional spoken translation — a browser-based, tool-agnostic option fills the gap Gemini leaves. You can read every word, in your language, on whatever platform the meeting happens to use.
Translate any conversation, on any platform
1 free hour to try. No credit card. No install. No bot in your meeting.
Get Started Free