Skip to content

Documents & scans

Scanned PDF to Word converter

Get editable document entities: headings, paragraphs, lists, tables, captions, page breaks, signatures, and visible form labels.

  • Scans detected automatically · AI Deep Read only when needed
  • Free: 5 conversions a day · 5 MB each
  • Deleted after 24 h · never used for training

Before you start

What to expect

  • Files are retained only for the product retention window and deleted after 24 hours unless your account policy says otherwise.
  • Uploads are not used to train models; AI Deep Read reads the page only to produce your requested output.
  • AI Deep Read costs 1 credit per rendered page and still needs human review for names, dates, legal text, and totals.
  • The DOCX is optimized for editing, not perfect visual fidelity to the scanned page.

What the output contains

  • A DOCX containing OCR text reflowed into editable paragraphs instead of a page image pasted into Word.
  • Headings, lists, simple tables, captions, and form labels reconstructed where AI Deep Read can identify them.
  • Page order preserved, with hard page breaks only where they help review the original scan.
  • Scanned PDFs are detected automatically; digital PDFs with a usable text layer can avoid Deep Read.

A real run

Scanned lease addendum

Input
9-page scanned PDF, 300 dpi, signature blocks, two small tables, and a stamped received date across page 1.
Output
DOCX with editable paragraphs, two recreated tables, signature lines as text, and one stamped date placed inline that needed manual repositioning.

Known failure modes and how they're handled

The PDF is only scanned page images
Deep Read runs automatically when no usable text layer is detected, producing editable DOCX text rather than image-only pages.
Faded, skewed, or low-resolution scans
The workflow extracts readable text but may misread faint characters. Review names, numbers, and legal clauses against the original.
Columns, sidebars, and stamps
Layout is reflowed for editing; complex columns may become a single reading order and stamps may appear as nearby text.
Password-protected PDFs
Unlock the PDF before upload. The workflow does not remove passwords or bypass restrictions.

Command line

Do it yourself

ocrmypdf

bash
ocrmypdf --deskew scan.pdf scan-ocr.pdf

Produces a searchable PDF text layer, not a reflowed Word document by itself.

pandoc --reference-doc

bash
pdftotext -layout scan-ocr.pdf scan.txt && pandoc scan.txt -o scan.docx --reference-doc=template.docx

Local pipeline is useful but requires manual cleanup for tables, reading order, and headings.

Questions about this workflow

Will the Word file still be an image?

No. The intent is editable DOCX text. Some signatures, stamps, or diagrams may be represented as text or omitted if they cannot be read.

Can it keep the exact scanned layout?

Not exactly. The output is reflowed for editing. Use a searchable-PDF OCR tool if preserving the original look is more important.

How do I know whether Deep Read will run?

The upload path checks for a text layer. Scans route to AI Deep Read; digital PDFs can use regular document extraction.

What happens to fillable PDF form fields?

Visible labels and entered values can be read, but interactive PDF form controls are not preserved as Word form fields.

← All workflows