How to make a scanned PDF searchable with OCR
A scanned PDF is only a set of pictures. You cannot search it, select a sentence or copy a number from it. OCR, short for optical character recognition, reads the letters in the picture and gives you real text. Here is how to use it and what affects the accuracy.
Step by step
- Open OCR PDF and add the scanned PDF.
- Choose the language of the document: English, Hindi, or English and Hindi together.
- Start the recognition. The first time, the language data is downloaded once; your document is not uploaded.
- Copy the extracted text from the box, or download the searchable PDF.
Getting accurate results
- Start with a clear scan: straight, evenly lit and in focus.
- Higher contrast helps. Black text on a white page works best.
- Scan at a reasonable resolution. Too low and letters blur together; very high gives slow results with little gain.
- Choose the right language. English text read as Hindi, or the other way round, comes out as nonsense.
What OCR struggles with
Handwriting, very small print, stamps over text, shadows and page curl near a book spine all reduce accuracy. Tables and columns may come out in a different order than you see them. Numbers and similar-looking letters (such as O and 0, or l and 1) are the usual mistakes, so always check important figures by eye.
What to do with the text
- Search and copy it into another document.
- Turn the PDF into an editable Word file with PDF to Word.
- Have it read aloud with Listen to PDF.
- Save the searchable PDF so you can find words in it later.
Make better scans in the first place
If you scan with a phone, Scan to PDF gives clean, upright pages with a bright background, which is exactly what OCR likes.
A worked example: a scanned Hindi and English notice
Many notices mix Hindi and English. If you pick English only, the Hindi lines come out as strings of odd symbols. If you pick Hindi only, the English words and numbers may suffer. Choosing the combined English plus Hindi option reads both, at the cost of more processing time. After recognition, scan through the result and check names, dates and reference numbers by eye, because those are the places where a single wrong character matters.
Preparing a poor scan before OCR
- Rotate pages so the text is horizontal. Sideways pages are read badly.
- Crop away dark borders and fingers at the edges of a phone photo.
- If the page is grey and dim, redo the scan in better light or use Scan to PDF with the Enhanced look, which brightens the paper and darkens the text.
- Photograph one page at a time, flat and square to the camera.
How to proofread quickly
Place the extracted text beside the page and compare the lines that carry facts: names, addresses, amounts, dates and ID numbers. Then run a spell check on the rest. It is faster to correct a few mistakes than to retype the page, but never assume that every character is right.