My scanner app says it saves a searchable PDF; can this tool get the text out of it?
Yes, if the app really added a text layer. The tool reads that layer like any other text, so you get words back. If it did not, the tool flags the pages as scanned and extracts nothing, because it does no OCR itself. Any mistakes in the words come from the scanner app's recognition, so proofread names and numbers.
What searchable PDF means
A scanner app takes a picture of the sheet, then tries to read the characters in it with optical character recognition. If it succeeds, it lays the words it read as an invisible text layer over the image, in roughly the same positions. You see the picture, but underneath the file carries text you can search and copy. That layer is what the PDF text extractor reads; it makes no distinction between text exported from Word and text placed by a scanner app.
How to tell whether the layer is there
The tool sorts each page with a simple rule: if it finds characters, it is a text page; if it finds no characters but does find an image, it is a scan.
An invoice photographed with an app that added a text layer containing Invoice 1042: classed as a text page, with the text and its character count shown.
The same page from an app that added no layer, or saved an image only: classed as scanned, with a notice that it has no text layer.
Some apps only add the text layer once a particular option is switched on, or with a specific export format. If every page comes back as scanned, check the app's export settings and save the file again.
The accuracy is not the tool's
The tool passes on what is in the layer exactly as it is and corrects nothing. If the app read an 8 as a B or a zero as the letter O, you get precisely that mistake. These places always deserve a check:
- Invoice, phone and bank account numbers, where one wrong character changes everything.
- Names, especially Arabic names that are handwritten or in a decorative font.
- Amounts and decimal points.
- Arabic text in general, since recognition quality varies widely from app to app.
Lines may come out less tidy
An OCR text layer is placed over the image approximately, and it can tilt if the sheet was photographed at an angle. Since the tool builds lines from the height of each piece of text, a tilted line may split in two, or a word may join the line next to it. Photograph the sheet flat and from directly above for a cleaner layer and better ordered text.
Common questions
Can the tool add a text layer to a scanned file?
No, it only reads what is there. Adding the layer is the job of OCR software or an app.
The extracted text has odd symbols in it. Why?
Most likely the app's recognition misread those spots. Compare them with the image in the original file.
Is the file uploaded to read the text layer?
No. All the reading happens in your browser.