Account

What new tool do you want?

PDF Tools

OCR PDF (Make Scanned PDFs Searchable)

Recognize the text in scanned PDFs and create a searchable PDF you can select and copy from — or extract the text — in 100+ languages.

How to use

  1. 1Add a scanned PDF and choose its language.
  2. 2Press Run OCR — you'll get a searchable PDF and the extracted text.

Frequently asked questions

How accurate is OCR?

Clean 300 DPI scans usually reach 95–99% accuracy. Blurry photos, handwriting and unusual fonts reduce accuracy.

Why is it slow?

Recognition runs on your device (Tesseract OCR in WebAssembly). Expect a few seconds per page.

Is my document uploaded?

No — only the OCR engine and language data are downloaded from a CDN; your file stays on your device.

What's actually running: Tesseract, in your browser

This tool downloads Tesseract OCR compiled to WebAssembly along with the language data you select (roughly 10-20 MB, fetched once from a CDN and cached) and runs text recognition entirely on your device — your scanned document itself is never uploaded anywhere, only the generic OCR engine and language files come down. Each page is rendered to a canvas at 2.5x scale before recognition, since OCR accuracy depends heavily on image resolution — a page rendered too small gives the engine fewer pixels per character to work with and measurably lowers accuracy.

What 'searchable' really means for the output file

The result isn't just extracted text sitting next to the original — each recognized page comes back from the OCR engine as a one-page PDF containing the original page image plus an invisible text layer positioned exactly behind the visible characters, which this tool then scales back to match the source page size and assembles into one output document. That's why the output looks pixel-identical to the scan but suddenly becomes selectable and searchable: the image you see hasn't changed, there's now an invisible, correctly-positioned text layer underneath it that a reader or search tool interacts with instead.

Why accuracy depends entirely on input quality, not the engine

The tool's own estimate is 95-99% accuracy on a clean 300 DPI scan, and that range is realistic for Tesseract specifically — but it drops substantially for blurry phone photos, low-contrast scans, unusual or decorative fonts, and handwriting, which OCR engines in general handle poorly regardless of which one you use. Selecting the correct document language before running matters too: Tesseract's language models are trained per-language, so running English-language recognition on a French or German document will misread accented characters and language-specific letter combinations even on an otherwise perfect scan.

Example: making a 30-page scanned contract archive searchable

A law office has a 30-page contract that only exists as a flatbed-scanned PDF — no selectable text anywhere, making it impossible to search for a specific clause across the document. Running OCR PDF with English selected processes roughly 3 minutes for 30 pages per the tool's own estimate (about 6 seconds per page), downloading the OCR engine once if it isn't already cached, and returns both a searchable PDF (visually identical to the scan, now with Ctrl-F working) and the raw extracted text split by page markers. If the original scan was done at a low resolution or at an angle, expect more misrecognized words concentrated in blurrier regions — worth a quick proofread of the extracted text against the original before relying on it for a legal search.

Related tools

More: PDF Tools