You try to copy a sentence from a PDF, but you cannot select it, or pasting gives you nothing. That PDF is most likely an image PDF made by scanning paper. The technology that pulls text out of such documents is OCR.
Text PDFs versus image PDFs
Text PDF
Created with "Save as PDF" in Word or Hangul. The file contains information such as "draw the character 'A' at this position", so you can search and copy it. The "Copy all text" feature of the PDF Viewer reads this information.
Image PDF (scanned PDF)
A PDF containing photos taken with a scanner or a phone camera. To a person it looks like text, but to the computer it is just a grid of pixels. Copying returns nothing.
What is OCR?
OCR (optical character recognition) analyzes the shapes of letters in an image and converts them into real text. It is now common enough to be built into operating systems, so you can get fairly accurate results without paid software.
Free ways to run OCR
Windows: PowerToys Text Extractor
Install Microsoft's free PowerToys utility, press Win + Shift + T and drag over any area of the screen to copy its text. It is handy for grabbing parts of a PDF shown on screen, and the processing happens on your PC.
macOS: Live Text
Open a scanned PDF or image in Preview, and you can select and copy text as if it were normal text (on macOS versions that support Live Text).
iPhone and Galaxy: camera and gallery
- iPhone: open the photo in the Photos app and press and hold the text to select and copy it.
- Galaxy: open the image in the Gallery and tap the text extraction button (the T icon).
Cloud document services
If you upload an image or PDF to Google Drive and open it with Google Docs, the OCR text appears with it. It is convenient, but remember that the file is uploaded to an external server. For sensitive documents, prefer OCR on your own device.
Tips for better OCR accuracy
- Resolution: scan at 300 dpi or more. Very small text is misread more often.
- Straightness: place the document straight. Skewed pages mix up lines.
- Lighting: avoid shadows and glare. Black text on a white background is most accurate.
- Tables and columns: the reading order can get mixed up in tables and two-column layouts, so extract them area by area.
Always check the result
OCR often confuses 0 and O, 1 and l, and similar-looking Korean syllables. Compare numbers (amounts, dates, account numbers) with the original in particular. The character counter and privacy masking tool are handy for checking the length of the extracted text and cleaning up personal information.