How OCR makes scanned PDFs usable
A scanned document looks like text, but to a computer it is only an image. You cannot search it, copy from it, or edit it. Optical character recognition (OCR) analyzes the shapes on the page and turns them back into real characters, so scans become searchable, editable and convertible.
Where OCR is used in these tools
PDF to Text recognizes the words on image-based pages so you can copy, search or download them. PDF to Word runs OCR on scanned pages so the resulting document is editable. PDF to Excel uses OCR to read tables in scanned statements and reports, and lets you choose the document's language before recognition.
Recognition runs in the browser using the open-source Tesseract engine. The first time it is used, the browser downloads the language data, typically 2–10 MB, which is then cached for later use.
What affects accuracy
OCR works best on sharp, straight, high-contrast scans of printed text. Scanning at 300 DPI is a good standard. Accuracy drops with skewed pages, shadows from phone photos, faint photocopies, decorative fonts, stamps over text and handwriting. Choosing the correct language matters because the engine uses it to decide between similar-looking characters.
Checking OCR output
Even good OCR makes occasional mistakes, typically confusing similar characters such as 0 and O or 1 and l. Proofread names, numbers, dates and amounts before relying on recognized text, especially for financial or legal documents.
Better scans, better OCR
- Scan in even light, flat and straight, at 300 DPI.
- Rotate sideways pages before running OCR.
- Select the document's language before recognition.