Private Translator

Translate text between English, Spanish and other languages with AI that runs in your browser, even offline.

Checking what this browser can run…

0 / 20,000
The translation appears here.

Downloaded language pairs

None yet. Each pair you download is listed here, ready to use offline.

Runs in your browser: your text never leaves your device. The first use of a pair downloads its model from Hugging Face and the AI engine (about 6 MB) from jsDelivr; after that it works offline. Models: OPUS-MT by the University of Helsinki (Jörg Tiedemann and Santhosh Thottingal, “OPUS-MT — Building open translation services for the World”, EAMT 2020), licensed CC-BY 4.0, converted to ONNX by Xenova. Engine: Transformers.js and ONNX Runtime.

Embed this tool on your site

Free and without an account: copy the code and paste it into your website, blog or docs.

About the Private Translator

Online translators send what you paste to their servers. This one runs open machine translation models inside your browser, so contracts, emails, medical notes or anything else you translate never leave your device. Pick the two languages, download that pair once (about 110 to 140 MB), and translate as much as you like, also without an internet connection.

How to use the Private Translator

  1. Choose the languages

    Pick the language of your text and the one to translate it into; the arrows button swaps them. The box above the text shows which model the pair uses, its size and whether it is already in your browser.

  2. Download the pair once

    Click Download and translate (or Download) to fetch the model from Hugging Face. It is kept in your browser, so the next time the pair is Ready (downloaded) and works offline. Downloaded language pairs lists what is stored and lets you delete it.

  3. Paste your text and translate

    Type or paste up to 20,000 characters and click Translate, or leave Translate as you type on. The translation appears sentence by sentence, keeps your line breaks, and Copy puts it on the clipboard.

When to use it

Confidential documents

Translate contracts, HR files, patient letters or internal emails without handing them to a translation service: the text is processed on your own device and nothing is uploaded or logged.

No connection

Download the pairs you need before a flight or a trip abroad and translate menus, messages and signs you type in, with no data plan or Wi-Fi.

Everyday reading and writing

Get the gist of an article or an email in French, German or Italian, or draft a reply in Spanish or English, then polish the wording yourself.

Frequently asked questions

Which models does it use, and under what license?

OPUS-MT, open translation models trained by the Language Technology Research Group at the University of Helsinki (Jörg Tiedemann and Santhosh Thottingal, EAMT 2020), in the ONNX conversion published by Xenova on Hugging Face. The OPUS-MT project releases its models under CC-BY 4.0 and the Hugging Face cards of most of these pairs say Apache 2.0; both allow commercial use with attribution, which the page gives below the tool. They run with Transformers.js and ONNX Runtime.

Which language pairs are there, and how big are they?

English, Spanish, French, German, Italian and Portuguese, in every direction: 30 combinations. Nineteen have their own model; the other eleven (Portuguese with anything but English, French to Italian, and German and Italian between each other) go through English with two models, which can lose details such as gender or formality. Each model is about 110 to 140 MB on WebAssembly and about twice that with WebGPU, which uses the 16-bit version; the page shows the exact size before downloading. Portuguese output is Brazilian Portuguese.

Is my text sent anywhere?

No. The translation runs in your browser, and neither the text nor the translation is sent to Tooly or anyone else. The only downloads are the model, from Hugging Face, and the AI engine (about 6 MB), from jsDelivr, the first time you use a pair; they contain nothing you typed. The last text and its translation are kept in this browser so they are still there when you come back.

Does it really work offline?

Yes, once a pair has been downloaded: the model stays in your browser storage and the engine in the site cache, so you can translate with no connection. If you are offline and choose a pair that is not downloaded yet, the page tells you to connect once to get it. Clearing your browser data, or deleting a pair from the list, removes the model.

How good is it compared with Google Translate or DeepL?

Good enough to understand a text and for plain, everyday writing, but less polished than the big cloud translators. These are compact models that translate one sentence at a time, so they miss context between sentences, stumble on idioms, slang and long complex sentences, and can get specialist terms wrong. Read the result before you send or publish it. Speed depends on your device: usually around a second per sentence or less on WebAssembly, and faster with WebGPU on a recent graphics card.