Scanned PDF to editable Word, page by page, in any script

Why scanned books trip up general AI assistants

A scanned PDF has no text inside, only pictures of pages. An assistant has to look at every picture and write out what it sees, and with hundreds of pages it tends to read the first few and work out the rest from context. That is a summary, not a transcription.

Scripts add a second problem. Urdu Nastaliq slopes and stacks letters, so dots and joins are easy to misread, and models trained mostly on Arabic and Persian tend to write Arabic ي or ك where Urdu needs ی or ک. Pashto has letters of its own, and Ge'ez writes a vowel as a small mark on each syllable. This tool reads each page on its own, at high resolution, with instructions for the script you choose, and asks the reader to mark uncertain words instead of guessing.

What people convert

• Scanned Urdu, Persian and Pashto books for research, re-publishing or translation
• Old family documents and letters typed or printed in Amharic, Tigrinya or Arabic
• University readers and theses that exist only as scans
• Court and immigration documents that must be translated, so they first need to be text
• Religious and literary texts for study, search and quotation

Limits, and getting a good result

Scans at 300 dpi or clear phone photos read best; blurred, very small or heavily damaged print may produce more highlighted words. Handwritten pages are a different job: use Handwriting to Text. Layout is rebuilt as paragraphs and tables, not as a pixel copy of the page: columns become running text, and images are left out.

Fonts in the Word file are chosen for the script (for example Urdu Typesetting for Urdu); if your computer lacks the font, Word substitutes one, which you can change with one click. One payment covers up to 300 pages in up to 50 files of up to 200 MB; Unlimited covers 600 pages per job. To translate the result, upload the Word file to Translate Word Document.