Make a scanned PDF searchable
Run OCR over every page so hidden text sits under each picture of a page. The book looks the same, and search, highlight and copy start working.
Key takeaways
- A scanned PDF is pictures of pages. Until OCR adds a text layer, Ctrl+F finds nothing and you can't highlight a word.
- OCR on a grey, speckled scan reads letters wrong. Cleaning the page to black text on white first gives the reader less to guess at.
- Many online tools cap OCR by page count: Sejda stops free users at 10 pages and paid users at 100.
- FixMyScan reads 19 languages, including English, German, French, Latin, Italian and Spanish, plus Fraktur and Ancient Greek, and picks the language itself.
- The file stays a PDF. Page order, pictures and page numbers don't move.
What “searchable” means in a scanned book
A scanner or phone camera saves each page as an image. To your computer, page 212 of a scanned history book is no different from a holiday photo: there are no letters in it, only pixels shaped like letters.
OCR (optical character recognition) looks at those shapes, works out the words, and writes them into the PDF as an invisible layer lined up with the printed lines. You still see the original page. Underneath it, the words now exist, so you can:
- search the whole book for a name, a place or a phrase
- highlight a passage and keep the highlight
- copy a quotation into your notes without retyping it
- tap a word for the dictionary on an e-reader
Test your PDF in ten seconds
Before you convert anything, find out what you have:
- Open the PDF on a computer.
- Pick a word you can clearly see on the page, say a chapter title.
- Press Ctrl+F (Cmd+F on a Mac) and search for it.
If the search lands on the word, the text layer is already there, though it may still be poor. If it finds nothing, the book is images only. Old downloads from Internet Archive, Google Books and HathiTrust can go either way, so test before you assume.
Why OCR on a raw scan reads nonsense
People who OCR a book once and get “tbe” for “the” and “rn” for “m” usually blame the OCR. More often the page itself is the problem. OCR software decides, pixel by pixel, what is ink and what is paper. A raw scan makes that hard:
- Grey or yellowed paper sits close in tone to faint print, so thin strokes vanish.
- Speckles and show-through from the other side of the leaf look like full stops and commas.
- A tilted page puts each line on a slope, and lines get read into each other.
- The dark gutter shadow near the spine swallows the first letters of every line.
So FixMyScan cleans first and reads second. Each page is levelled to white, the print turned to sharp black, specks removed and small tilts straightened, and only then does the OCR run over it. The cleaned page is also what you read, so the text you search lines up with the page you see.
How to make a scanned book searchable with FixMyScan
- Upload the book. A scanned PDF or a DjVu file, up to 1,000 pages. You don’t pick a language.
- Set the crop once. The pages are shown stacked on top of each other, and one box trims the margins of the whole book.
- Let it run. Every page is cleaned, then read. Pages with photos or engravings are kept in greyscale rather than forced to black and white.
- Check the flagged pages. Pages that came out blank, cropped oddly, unsure in the reading or full of pictures are marked. Swap any page back to the original if you prefer it.
- Download the PDF. Search it on your computer, or send it to your e-reader.
Languages and old type
The language is found from a few sample pages, so a mixed book doesn’t need a setting. FixMyScan reads:
| Script | Covered |
|---|---|
| English, German, French, Italian, Spanish | Yes |
| Latin | Yes |
| Fraktur (old German blackletter) | Yes, found automatically |
| Ancient Greek, including Greek quoted in a Latin or English text | Yes |
| Handwriting | No |
| Right-to-left scripts, Chinese, Japanese | Not covered |
The reading engine underneath is Tesseract, whose language data includes Fraktur, Latin and Ancient Greek models.
Long books and page caps
The most common complaint about online OCR is a book that is too long for it. Here is what the tool sites themselves state:
- Sejda: free users get 10 pages per OCR task, paid users documents up to 100 pages.
- iLovePDF: free OCR is 1 task at up to 15 MB.
- Adobe’s online OCR: free, but you sign in to download, and it states no page cap.
A 400-page scanned textbook doesn’t fit the first two without cutting it into pieces, running each one, and stitching the results back together. FixMyScan takes the book whole, up to 1,000 pages.
If you are comfortable with a terminal, OCRmyPDF is a free option with no page cap. It adds a text layer to the original PDF and can straighten pages, but it doesn’t crop margins or whiten the paper, and on Windows you install Tesseract and Ghostscript as well.
What you can do with the searchable copy
Researchers search a 600-page county history for one surname and land on every mention. Students copy a definition straight into their notes. On a Kindle, Kobo or tablet, a long-press now selects a word, which means highlights and dictionary look-ups work on a book that was only pictures before.
Page numbers stay where they were, so a quotation you copy can still be cited to the printed page.
What OCR won’t do
It won’t turn the book into an EPUB that reflows to any font size. The pages stay pages. If you want to know why that is usually the better outcome for a scanned book, see scanned PDF or EPUB.
Questions
How do I know if my PDF is already searchable?
Open it on a computer, press Ctrl+F (Cmd+F on a Mac) and search for a word you can see on the page. If nothing is found, or you can't drag to select a line, the PDF is pictures only. Our free scanned PDF check tells you in the browser without uploading.
Does OCR change how the pages look?
OCR on its own doesn't: the text layer is invisible. FixMyScan also cleans the pages before reading them, so you get white paper and black text as well. Pages with pictures stay greyscale.
Is PDF OCR free?
It can be. OCRmyPDF is free and open source, but it runs from the command line. FixMyScan costs $9.99 a week or $79.99 a year.
Will search work on my Kindle or Kobo afterwards?
Yes. The text layer is part of the PDF, so any reader that searches PDFs can use it, on an e-reader, a tablet or a computer.
Can it read handwriting?
No. It reads printed type. For handwritten letters or registers, a handwriting tool such as Transkribus is built for that.
What about a book in Fraktur or with Greek quotations?
Both are covered. Old German blackletter is read as Fraktur, and Ancient Greek is added when enough of the sampled words are Greek, so a classics text with Latin and Greek on the same page is read as both.
Is OCR text ever wrong?
Yes, on any tool. Faded print, tight bindings and broken letters all cause misreads. FixMyScan flags pages where the reading looked unsure, so you know which ones to check.
What happens to my file?
It stays private to your account. Every upload is deleted 24 hours after the book is ready, or 24 hours after upload if you don't convert it.