How to Make a Scanned PDF Searchable With OCR, Right in Your Browser
You have a scanned 40-page rental agreement and need to find the clause about the security deposit. Ctrl+F finds nothing, because a scan is a photograph of paper and there are no letters in it. OCR PDF fixes that: it reads the words on each page and lays them back over the page as invisible text, so search, select and copy start working and the page still looks exactly the same.
What's going on under the hood
Each page is rendered at 300 DPI and handed to Tesseract, the long-standing open-source OCR engine, here compiled to WebAssembly so it runs in the browser. Tesseract reports each word and where it sits. Those words are then written into the original page as a hidden text layer at the same positions. The image itself isn't redrawn or recompressed.
If some pages already contain real text (say, a typed cover page followed by scans), they're skipped by default. You can untick that option to force OCR on everything.
How to run it
- Open the tool and choose the scanned PDF.
- Choose the language: English, Hindi, or English and Hindi for mixed pages.
- Optionally set a page range if you only care about part of the file.
- Press Start OCR. Each page shows its own progress; Cancel stops straight away and saves nothing.
- Download the searchable PDF. The recognised text also appears below the button if you just want to copy it.
Example: a Hindi-English government letter
Lots of Indian official documents mix Devanagari and English on the same page. Pick English and Hindi and Tesseract handles both scripts in one pass; the Hindi words are embedded with a Devanagari font. One quirk the tool page is upfront about: some PDF readers copy words containing the vowel sign ि with the letters in a different order. If you need exact Hindi text, copy from the text box on the page rather than from the PDF.
Getting better results
- Start with a decent scan. Clean printed text at a sensible resolution comes out well. Blurry phone photos, handwriting and tiny print cause errors.
- Straighten crooked photos first. A heavily tilted page confuses line detection. Scan to PDF can help you capture straighter pages in the first place.
- Check numbers by eye. OCR can mistake 0 for O or 1 for l. For amounts, dates and account numbers, compare against the image.
- Limit the range on big files. OCR is heavy work for a phone. Process just the pages you need.
- Compress after, not before. If you later shrink the file with Compress PDF, use its Keep text and links mode; the picture modes would flatten your new text layer back into an image.
What it won't do
It doesn't turn the scan into an editable Word document by itself. For that, OCR first and then run the result through PDF to Word. It also isn't magic on handwriting; treat any handwritten page as "might work".
Privacy
This is the part I care most about for OCR. Cloud OCR services see every word of your document. Here, the tool page states the OCR runs on your device and the page and its text aren't uploaded or stored, and the engine files come from the site itself. You can confirm it in your browser's network tab. The page shows Google ads like the rest of the site.
FAQ
Will the pages look any different? No. The original image stays; only an invisible text layer is added.
How accurate is it? Very good on clean printed pages, shaky on blurry photos and handwriting. Check anything important.
Does it read Hindi? Yes, alone or together with English.
Can I stop halfway? Yes. Cancel stops within about a second and nothing is saved.
Have a scan you can't search? Run it through OCR PDF.
— Alok Kumar Singh