What OCR PDF does
When you scan a contract, a receipt or a page of a book, the scanner saves a photo of the page. To your computer it is just pixels. You cannot search it, select a sentence or copy a number out of it, and screen readers cannot read it aloud.
OCR, short for optical character recognition, fixes that. It looks at the shapes in the image, works out which letters and words they are, and writes them into the PDF as an invisible text layer that sits exactly behind the printed words. The page looks the same as before. The difference shows up when you press Ctrl+F, drag to select a line, or paste a quote into an email.
How to make a scanned PDF searchable
- Choose your PDF. Any scanned or photographed PDF works, including files from phone scanner apps and office copiers.
- Pick the language the document is written in. If it mixes in English words, leave “Also read English words” ticked.
- Choose the accuracy. Best renders each page at 300 dpi, which is what OCR engines like for normal-sized print. Faster uses 200 dpi and suits big, clear text.
- Click “Make pages searchable”. Each page is drawn, cleaned up and read in turn. A progress bar shows which page is being read.
- Download the searchable PDF. You also get all the recognised text on screen, ready to copy or save as a .txt file.
How it works inside your browser
The tool uses three open-source pieces, all running on your device:
- PDF.js draws each page as a high-resolution image.
- Tesseract, a mature OCR engine first developed at HP and later at Google, reads the image. It runs as WebAssembly in a background worker, so the page stays responsive.
- pdf-lib places Tesseract’s invisible text page over your original page.
Your original page is not replaced or recompressed. The scan keeps its exact quality, and the text layer is drawn in “invisible” mode (PDF text rendering mode 3), which is how professional OCR software does it. The text layer adds only a few kilobytes per page.
The first time you use a language, its data file (between 0.4 and 3.7 MB) downloads from jsDelivr, a public code CDN, and your browser caches it. Only that data is downloaded. Your PDF is never uploaded, and you can check it: open your browser’s developer tools, go to the Network tab and run the tool. No request carries your file.
Getting the best accuracy
OCR quality depends far more on the scan than on the software. These habits make the biggest difference:
| Do this | Why it helps |
|---|---|
| Scan at 300 dpi | Letters need roughly 20 to 30 pixels of height to be read reliably |
| Keep pages straight | Tilted lines confuse line detection; most scanner apps have an auto-straighten option |
| Use good, even light for phone photos | Shadows across the page turn into grey smudges |
| Choose the right language | Each language model knows its own letters, accents and common words |
| Leave “Boost contrast” on for faded copies | It stretches grey text towards black and the paper towards white |
Expect very good results on printed letters, invoices, forms and books. Handwriting, decorative fonts, text on busy backgrounds and tiny footnotes are much harder. The tool shows an average confidence score for the document. Below about 70%, check the text before you rely on it.
A worked example
A 14-page scanned lease arrives as a 9 MB PDF. Searching it for “deposit” finds nothing, because the pages are pictures. With the language set to English and accuracy on Best, a recent laptop reads it in about one minute. The result is about 9.1 MB, almost the same size, because only text was added. Searching for “deposit” now jumps to clause 4.2 on page 3, and you can copy the clause straight into an email.
Mixed documents
Many PDFs are partly digital and partly scanned. For example, a typed report may have signed scan pages at the end. With “Skip pages that already have selectable text” ticked, the tool checks each page first and only reads the ones without text. That saves time and avoids putting a second, duplicate text layer on top of real text.
Free OCR vs typical online OCR services
| This tool | Typical online OCR services | |
|---|---|---|
| Where your document goes | Stays on your device | Uploaded and processed on a server |
| Files per day | Unlimited | Often one free file, then a paid plan |
| Pages per file | Unlimited | Commonly capped on free plans |
| Account required | No | Often, for larger files |
| Speed | Depends on your device | Depends on the upload and the server queue |
| Languages | 39, switchable per file | Varies |
Server tools can be faster on very long files, because their computers are powerful. The price is that a copy of your document (often an ID, a contract or a medical letter) sits on someone else’s server for a while. Here it never leaves your computer or phone.
Limits worth knowing
- Speed: OCR is heavy work. Allow two to five seconds per page on a laptop and longer on an older phone. Very long books are better done in parts with Split PDF.
- Upside-down or sideways pages: rotate them first with Rotate PDF, because the reader expects upright text.
- Layout: the text layer follows the words’ positions, but the .txt download is plain text without columns or tables.
- Password-protected PDFs must be unlocked first with Unlock PDF (you need to know the password).
What to do next
A searchable PDF opens up other tools. Convert it into an editable document with PDF to Word, pull out just the words with PDF to Text, or make it smaller for email with Compress PDF. For a single photo or screenshot rather than a PDF, use Image to Text.