OCR a scanned PDF

A scan is a picture of a page: there is no text in the file to select, search or copy. OCR reads the words from the pixels and writes them back as an invisible layer sitting exactly over the printed ones, so the page looks identical but its text can be found and copied. The recognition and the language model both run on your device.

How to do it

1. Open the scanned PDF.
2. Open the OCR panel and start.
3. Each page is read at 300 DPI; the first run also loads the language model.
4. Save the searchable copy.

Why nothing is uploaded

Most online PDF tools work by sending your file to a server, editing it there, and sending it back. That means contracts, medical records, bank statements and passports all end up on someone else's machine, and you are trusting a privacy policy about what happens next.

This editor does the work in your browser instead. The PDF is read by this tab, rendered with pdf.js and written back out with pdf-lib, both running as ordinary JavaScript on your own device. There is no upload endpoint, because there is no server doing any of this. Once the page has loaded you can disconnect from the network entirely and everything still works.

Questions

Is my file uploaded?

No. Everything happens in this browser tab, on your own device. There is no upload endpoint and no server-side processing, so the file never leaves your machine — you can disconnect from the network after the page loads and it still works.

Does the language model come from a third party?

No. Both the recognition engine and the English model are served from this site, not a CDN — a tool promising no third-party requests should not quietly make one the moment you use a feature. The model is about 3 MB and your browser caches it after the first run.

How accurate is it?

Good on clean, straight scans of ordinary printed text; worse on handwriting, low resolution, heavy skew or unusual typefaces. Words the engine is not confident about are left out rather than filling your document with plausible-looking nonsense.

Does the page look any different afterwards?

No. The recognised text is drawn in an invisible rendering mode over the original image, so nothing is painted — you see the scan exactly as before, but selecting and searching now work.

Related tools

• Extract text from a PDF — Pull the text out of a PDF as plain text or Markdown, without uploading it anywhere.
• Compress a PDF — Make a PDF smaller in your browser, with an honest account of what each option costs. Nothing is uploaded.
• Redact a PDF — Black out text so it is genuinely gone, not merely covered. Redacted pages are flattened to images on export, in your browser.
• The full editor — annotate, sign, fill forms, redact, OCR and more.

Part of messy-ui. Free, open source and account-free — see also the countdown timer.