Getting the most out of PDFs
How do I translate a PDF without sending it to a server?
Recent Chrome and Edge include an on-device translator a web page can use, so a PDF's text can be translated with nothing uploaded — at the cost of a one-time language-pack download, and with real gaps in which language pairs are supported.
Translating a document usually means handing the whole of it to a translation service. For a menu that is fine. For a contract, a medical letter or anything with a name and an address in it, it is a decision worth making deliberately rather than by default.
There is now a way to avoid it entirely.
The browser has a translator in it
Recent versions of Chrome and Edge ship the same translation engine that powers "translate this page", and expose it to web pages. The model runs on your machine. The text never goes to a server — not to Google's, not to ours, not to anyone's.
Translate PDF reads your PDF's text in the browser, hands it to that engine paragraph by paragraph, and shows the original beside the translation so you can check it rather than trust it.
What it costs
The first time you use a language pair, the browser downloads a language pack. This is a real download and it is not small — expect to wait, on a slow connection for several minutes. It happens once. After that the pair works instantly and offline.
While it is downloading the page says so, with a percentage, because a blank screen for four minutes is indistinguishable from a broken feature.
Which browsers
- Chrome and Edge, version 138 or newer — supported.
- Firefox and Safari — no equivalent a web page can call.
On an unsupported browser the tool says so and stops. It does not quietly fall back to uploading your document somewhere, which would defeat the entire point of choosing it.
Which languages — and an honest gap
The engine supports a good range of major languages, and the quality on the common European pairs is genuinely good.
It does not currently support English to Bangla. The browser reports the pair as available, downloads a language pack for it, and then refuses to build the translator. That is a limitation of the browser's engine, not of this tool, and there is nothing a web page can do about it. If you need Bangla translation today, this is not the tool — and it says so directly rather than failing after a five-minute wait.
That gap may close: the browser's language support grows with each release, and the tool checks per pair every time rather than assuming.
Before you translate: is there text?
The same rule as every text tool. A scanned page is an image of words and there is nothing to translate. OCR PDF turns the picture back into text first.
What you get back
Paragraphs, with the original beside the translation. Not a translated PDF with the original layout — that is a much harder problem, because translated text is a different length from the original and reflowing a laid-out page around it changes the document. What you get is the text, in order, with the pages marked, which you can copy or download.
If you need the translation *in* the document, translate first, then type it in with the PDF editor where you control exactly where it goes.
Common questions
Does my PDF get uploaded?
No. The file is read by your browser and the translation is done by the browser's own engine on your machine. Nothing is sent to a translation service.
Why is it downloading something?
The first use of a language pair fetches that pair's model, which the browser manages and reuses. It happens once; after that the pair works offline.
It says my browser cannot do this — why?
Firefox and Safari have no on-device translator a page can use, and older Chrome and Edge do not either. The alternative would be uploading your document, which this tool will not do without telling you.
Can it translate English to Bangla?
Not at present. The browser advertises the pair and then refuses to create the translator for it. The tool reports that plainly instead of leaving you waiting.