Understand and be understood, in 100+ languages
Real-time voice, camera and screen translation — face to face or on the go. On-device by default, with optional cloud AI.
Real-time voice, camera and screen translation — face to face or on the go. On-device by default, with optional cloud AI.
The short version: your phone listens, reads and translates while the conversation is still happening — and most of that work never leaves the device.
Most translation tools work in batches. You finish speaking, you press a button, you wait for a result, and the other person waits with you. Real-time translation streams instead: words appear and correct themselves as the sentence forms, so you are already reading the second half of someone's thought while they are still saying it. That single difference decides whether a conversation flows or turns into a series of pauses, and it is why people describe a real-time translator less as a tool they operate and more as a shared language they can both step into.
Speech recognition and camera text recognition run on your own device rather than on someone else's server. That has three practical consequences. The core of Live Translator keeps working when you have no signal — on a plane, underground, or abroad without a data plan. Your conversations are not uploaded anywhere by default. And there is no per-minute meter running while you talk. Cloud AI is available when you want higher-quality translation, summaries or chat, but it stays opt-in: you decide whether to switch it on, and you can bring your own API key when you do.
Voice mode puts the phone between two people, splits the screen, and gives each side their own language to speak and read, so nobody has to pass the device back and forth. Camera mode translates the text in front of you — a menu, a street sign, a form, a page of a contract — either from the live view or from a photo you already took. Screen mode floats a translation over whatever app you are using, so a message thread, a listing or a website in an unfamiliar language becomes readable without copying anything out of it.
Most engines quietly assume one language per sentence. Real speech does not work that way. In Hong Kong a sentence like "我想去 Causeway Bay 開會" is completely ordinary, and generic models either mangle the English fragment or drop it entirely. Traverba is tuned for exactly that mid-sentence switching between Cantonese and English, which is the failure people report most often with mainstream translation apps. The same problem shows up wherever people mix languages naturally — Singlish, Taglish, Hinglish, Spanglish — and the fix is the same: treat code-switching as normal speech rather than as an error to be cleaned up.
A few habits make a noticeable difference. Choose the two languages up front instead of relying on auto-detect when you already know them, because detection costs a beat at the start of every exchange. Keep the phone nearer to whoever is speaking than to the room. Download the language packs you expect to need before you travel, while you still have decent bandwidth to spare. And when you are somewhere genuinely loud, switch to camera or screen mode — reading a translated menu is far more reliable than pulling clean speech out of a crowded restaurant.
Live Translator is built for the conversations that would otherwise not happen at all: asking for directions, ordering food, a parent-teacher meeting, a shift handover, a customer at the counter, a neighbour you have never quite been able to talk to. It is not a replacement for a certified interpreter in a courtroom, a hospital consultation or a contract negotiation, where a mistranslated clause carries real consequences and a qualified human is professionally accountable for the wording. Knowing which of those two situations you are in matters far more than any feature list.
A real-time translator turns speech or text into another language as it happens, rather than after the fact. Traverba does it for voice, camera and screen in 100+ languages, so you can follow a conversation while it is still going.
Yes. Put the phone between you and each person speaks and reads their own language on their own half of a split screen — no passing the device back and forth.
It covers the same need in everyday situations — travel, appointments, front-desk conversations — without booking anyone in advance. For legal, medical or diplomatic settings, a certified human interpreter is still the right choice.
Speech recognition and camera text recognition run on your device, so the core translation works with no connection. Cloud AI is optional and only used if you turn it on.
Over 100 languages, including mid-sentence Cantonese ⇄ English code-switching that most translation apps get wrong.