Make a scanned PDF searchable (OCR)
Add an invisible text layer to a scanned PDF so it can be searched and copied. Runs on your device.
Runs on your device. The file is never uploaded.
Make a scanned PDF searchable draws an invisible text layer over each page, placing every word where Tesseract read it. The 7MB English model downloads only after you tick the box, and this app keeps no copy of it. Pages that already carry text are skipped. The layer is Helvetica, so any character outside WinAnsi is dropped.
Questions
What does OCR add to my PDF?
An invisible text layer, drawn in the position where each word was recognised. The page still looks exactly as it did, you are seeing the original scan, but the text underneath can be searched, selected and copied.
Do I have to download anything?
Yes, once: the English OCR model, which the page asks you to confirm before it starts. It is served from this site rather than a third party, and this app stores it nowhere, your browser's own HTTP cache answers the repeat requests, which is why a second run costs no bytes.
Does it re-OCR pages that already have text?
No. Pages that already carry a text layer are left alone and reported separately, so a mostly-digital PDF with a few scanned inserts only pays for the inserts.
What does the resolution setting change?
How large the page is rendered before it is read: 1 means 72dpi, 2 means 144dpi, and so on. Higher catches small print at the cost of time and memory; 2 is a sensible default for a typical scan.
How accurate is it?
It is Tesseract, so: accurate on clean, straight, printed English; worse on handwriting, heavy skew, low contrast or unusual fonts. Words it was unsure about are reported, so you know which pages to check rather than trusting it blindly.
Why does this one need a browser?
Because rendering a PDF page to an image needs a canvas, which Node does not have. The recognition itself runs anywhere, the plain image OCR tool does run on the command line, but this tool has to draw the page first.