When you need OCR, and when you do not

There is a one second test for any PDF. Open it and try to select a word with the cursor. If the text highlights, the document already contains real text, and recognition has nothing to add. If the whole page highlights as one block, or nothing happens, you are looking at a picture and the words are not text at all.

That second case is what OCR is for. It reads the shapes in the picture and writes a text layer underneath, leaving the picture itself on top. The page looks exactly as it did, but the words can now be searched, selected and copied, and other tools can finally work with the content.

Doing this to a document that already has text is a waste at best. Recognition would read the rendered image of text that was already perfect and replace certainty with a good guess, usually while making the file larger. Our tool skips pages that already carry text for this reason.

OCR is also the missing step behind several conversions that people expect to fail. A scanned PDF converted to Word or Excel gives you a picture pasted into a document, because there was nothing else to give. Recognise it first, then convert, and the result is the editable file you were expecting.