Can You Translate a Scanned PDF? Yes — Here's How

Yes — a scanned PDF can absolutely be translated. The catch is that a scan contains no actual text, only a picture of text, so it has to go through OCR (optical character recognition) before any translation can happen. Every workflow that includes that step works, from free manual methods to one-click tools. Every workflow that skips it fails — usually with an error message, a blank result, or garbled output.

If you've already tried and hit one of those failures, nothing is wrong with your document. This guide explains why so many tools reject scans, walks through the three realistic ways to translate one, sets honest expectations for quality, and covers the cases where machine translation isn't enough on its own.

Why it looks like you can't: the text-layer problem

A PDF exported from Word, a browser, or an invoicing system carries an invisible text layer — the actual characters, sitting underneath what you see. Software can select, copy, search, and translate that layer directly. A scanned PDF has no such layer. It's a photograph of a piece of paper wrapped in a PDF container, and to most translation tools a photograph is unreadable.

There's a ten-second way to check which kind you have: open the PDF and try to select a few words with your cursor. If individual words highlight, there's a text layer and most translators will accept the file as-is. If your cursor drags an empty box across the page, you're looking at an image — and you need a workflow that includes OCR.

The three realistic ways to translate a scanned PDF

1. OCR it yourself, then translate the text

Run the scan through an OCR tool — many free web tools exist, and most desktop PDF apps with OCR can produce a searchable file — then copy the recognized text into any translator you like. This is free and it genuinely works for understanding what a document says. What it doesn't give you is a document: the moment you copy text out of the page, the layout is gone. A form becomes a list of disconnected labels, a table becomes a run-on paragraph, and you rebuild any structure yourself. Fine for a one-page letter; slow and error-prone for anything with tables, columns, or many pages.

2. Use an all-in-one PDF suite

Some paid PDF suites bundle OCR and translation together. If you already own one, this can be reasonable for simple documents. The friction people commonly hit: the translate feature only operates on a text layer, so you still have to run the OCR step first, and the output typically re-flows the page rather than reproducing it — so structured documents need manual formatting cleanup afterwards. Budget time for that cleanup on anything that isn't plain running text.

3. Use a translator built for scans

Purpose-built tools handle the whole chain in one pass, and they treat the layout as part of the job rather than an afterthought. Reglyph is one of these: OCR reads the page, the original text is erased from the page image, and the translation is typeset back in its place — so tables, stamps, figures, columns, and numbers stay exactly where they were. You upload the scanned PDF (or a photo of the page) and download a translated PDF that still looks like the original document. It works with 12+ languages, offers a bilingual side-by-side export for checking the result against the source, and lets you translate 5 pages free with no credit card; paid use starts at $5. Because it's a web app, it also works from a phone's browser.

What quality should you expect?

Two things determine the result: the quality of the scan and the nature of the content. A clean, flat scan of printed text — a laser-printed contract, a typed certificate, a book page — will OCR with very few errors, and modern machine translation on top of that is good enough that you'll understand the document completely and accurately in most cases. Low-resolution scans, faxes, skewed photos, and pages with heavy background noise recognize less reliably. Heavy handwriting is the weak point of every OCR system: a handwritten note or a fully handwritten form is unreliable with any automated tool, and you should expect to verify that content by other means.

Whatever tool you use, spot-check the parts of the output where a small error matters most:

• Numbers, dates, and amounts — a misread digit survives translation untouched, so compare them against the original image.
• Proper names — of people, companies, and places, which OCR can garble and translators sometimes 'translate' when they shouldn't.
• Legal or technical terms of art — machine translation renders the meaning, but a field may have one specific accepted term.
• Anything handwritten — signatures, margin notes, filled-in blanks on a printed form.

When you need a human translator

Machine translation output, however good, is not a certified translation. If you're submitting a document to an authority that requires certification — USCIS and many other immigration offices, courts, and some universities — a human translator must review the translation and sign a certification of accuracy. No software output alone satisfies that requirement, and any tool claiming otherwise is misleading you.

That doesn't make machine translation useless for these documents — it changes its role. Getting an accurate draft with the layout already reproduced means the translator (or you, before hiring one) starts from a nearly finished document instead of a blank page, which is typically faster and cheaper than commissioning a from-scratch translation. Do the machine pass first, then bring in the human for review and certification.

Bottom line

You can translate a scanned PDF — the only real question is how much of the work you want to do by hand. If you just need to know what one page says, free OCR plus any translator gets you there. If you need a translated document you can keep, share, or print, use a tool that runs OCR, translation, and re-typesetting in one pass. And if the destination is an official authority, use the machine output as the draft and have a human translator certify it.