When to run text recognition on a scan
Looking for one figure in a scanned contract means reading the whole thing with your eyes. A scan is a photograph of a page — as far as a computer is concerned it contains no text at all, so nothing is searchable and nothing can be copied.
This tool reads the characters out of the picture and lays them underneath as an invisible text layer. The page looks exactly as before, but from then on Ctrl+F finds things and you can select and copy. The file goes to our conversion server and is deleted the moment it finishes.
How to use it
- 1Upload the scanned PDF. Up to 30MB.
- 2Press Recognise text — the server reads it page by page. A long document can take a few minutes.
- 3Open the result and try searching or selecting the text.
Common questions
- How accurate is it?
- A Korean quote was tested under several conditions. A clean scan read perfectly, and so did a version degraded to look like a phone photo — two degrees of skew, speckle, blur and heavy JPEG compression — with all six lines exact. Those are favourable conditions, though: a photo with shadows across it, or a curled page, or a faded till receipt will do worse.
- Does the page look different afterwards?
- No. The original image is left alone and the recognised characters go underneath it, invisible. Nothing changes on screen until you drag across the text and a selection appears.
- Can I convert it to Word afterwards?
- Not yet. The tool behind PDF to Word is built to skip invisible text layers, so it comes back empty. PDF to Excel does read the layer — if the scan is a table, continue that way instead.
- Which languages does it read?
- Korean and English, including English company names and figures mixed into a Korean document. Japanese and Chinese are not supported.
- Can I run it on a PDF that already has text?
- You can, but you shouldn't. It throws away the real text, turns the page into a picture and reads it back — so exact characters become recognised guesses and the file grows. If a PDF is already searchable, leave it alone.