discreetpdf.com

Make a scanned PDF searchable with OCR

Drop a scanned PDF and get the same document back, now searchable and selectable. Convert a single file and you also get the recognized text as a .txt download. Recognition runs on your device.

Recognizes printed English text. More languages are planned.

Files are processed entirely in your browser. Your file's contents never leave your device.

Your PDF never touches a server. There's nothing here to ration.

No daily caps. No spinner that turns into a paywall after the third file.

Private by architecture

Your PDF never leaves your device.

Unlimited use

No daily caps or hourly throttling.

No signup, no watermarks

Every tool below works with or without an account.

Works offline

A tool you've already loaded keeps working offline.

About this tool

A scanned PDF is a stack of photographs. You can read it, but you can't search it, select a sentence, or copy a paragraph into an email. OCR fixes that. Drop the scan and you get the same document back, visually identical, with an invisible text layer lined up over the print. Ctrl+F finds things, selection works, copy-paste works. You can also download everything it read as a plain .txt.

The pipeline runs on your machine: PDFium renders each page at 300 DPI, the resolution OCR engines are happiest with, and Tesseract (the open-source engine Google maintained for years) does the recognition as WebAssembly in your browser. Even the language model loads from this site rather than a CDN, so every request stays same-origin and none carries your document.

Fair warning on accuracy: clean printed English does very well. Contracts, books, letters, invoices. Low-resolution scans, skewed photos, tiny print, and decorative fonts do worse, and handwriting isn't supported at all. OCR is guesswork at heart, and the occasional wrong word is normal. The scan itself stays on top, though, so what you see is always the true page. Only the hidden search layer can be wrong. English only for now; more languages are planned.

Frequently asked questions

Is my scan uploaded for processing?
No. The renderer, the OCR engine, and the language data all load from this site and run in your browser. Watch the network panel while it works: every request stays same-origin and none carries your document.
What does a "searchable PDF" mean?
Your pages stay visually identical, and an invisible text layer gets positioned over the print. Readers then treat the file like any digital document. Search finds words, selection works, copy-paste works, and you still see the original scan.
How accurate is the recognition?
On clean printed English at normal sizes, near-perfect with occasional misses. Accuracy drops with low-quality scans, skew, small print, and decorative fonts. Handwriting needs a different kind of model entirely and is not supported.
Which languages are supported?
English for now. The engine itself can do 100+ languages, and more are planned. Each adds a few MB of model data, so they'll be opt-in rather than always loaded.
How long does it take?
One to three seconds per page on a typical laptop, after an engine load of a few seconds the first time. A 50-page scan lands in a minute or two, with progress shown page by page.
My PDF already has some text. Can I still OCR it?
Yes. The tool warns you when it finds an existing text layer, since OCR on a digital PDF is usually pointless, but it will still run. Useful for mixed documents where some pages are scans and some aren't.

All PDF tools

All free, all in your browser. Pick the task you need.