Recover the text of a scanned PDF (OCR)

Prepare the document

The result depends on the image: 300 dpi or more, a straight page, sharp contrast. A phone photo benefits from a pass through “Fix a photographed document” first (straightening, black-and-white filter).

Run the recognition

With the OCR tool:

• Drop the image or scanned PDF and pick the language (French, English): the matching model (6 to 11 MB) is downloaded once and kept.
• Start the recognition; the orientation of each page is detected, even for rotated pages.
• Get the result: plain text to copy, or a searchable PDF where the invisible text is aligned with the original image.

Proofread and correct

OCR makes mistakes on pale scans, fancy fonts, tables and numbers: proofread the important passages. The average confidence shown gives an idea. In the PDF editor, the built-in OCR makes every recognised passage editable directly on the page.

To get an editable Word document, then convert the searchable PDF with “Convert a PDF to Word”.