How to convert a scanned Arabic PDF to Word
By PDFEditor.ae · Published
How do I convert a scanned Arabic PDF to Word?
Open PDF to Word and upload the scan. Set Document language to Arabic (or Arabic + English if the pages mix both), keep On this device (private) and click Recognise text. Correct the text in the preview, click Create searchable PDF and Word, then Download editable Word. The Word file holds the text as right-to-left paragraphs; it does not rebuild the layout, tables or images.
Why a scanned PDF needs text recognition first
A scanner or phone camera saves a picture of each page. The Arabic you see is pixels, not letters, so a normal converter finds nothing to put in Word. Text recognition (OCR) reads the picture and turns it back into characters. The tool checks every page when you upload and tells you how many need recognition. Pages that already have selectable text are reused as they are.
Step by step
- Work on a copy of the scan, not the only original.
- Upload it to PDF to Word and wait while the pages are checked.
- Under Document language choose Arabic, English or Arabic + English. Choose the language that is really on the page: Arabic + English is best for bilingual forms and invoices.
- Keep On this device (private). The first run downloads the Arabic language data (up to about 8 MB), then pages are read in your browser and not sent anywhere for recognition.
- Click Recognise text. Up to 50 pages can be recognised in one run.
- Read the text under Recognised text (you can edit it). Fix names, numbers, dates and amounts against the page.
- Click Create searchable PDF and Word, then Download editable Word. Download searchable PDF as well if you want a copy of the scan you can search and copy from.
What accuracy to expect
In our own tests on synthetic Arabic scans, on-device recognition misread between about 1 and 6 characters in every 100. A blurred, noisy and slightly skewed 150 dpi test scan dropped to about 1 in 100 after Retry with clean-up. Those test pages were printed text; real phone photos, stamps, handwriting, tables and faint copies will do worse.
If a page comes back with little usable text, try Retry with clean-up. Enhanced (more accurate) recognition is optional: it sends an image of each selected page to OpenAI, only after you tick the consent box, and there is a daily page limit.
Common problems
- Digits look wrong: scans may use Arabic-Indic digits (٠١٢٣) or Western digits (0123). Check every number against the page before relying on it.
- English words appear in odd places: choose Arabic + English rather than Arabic, so mixed lines are kept in reading order.
- The Word file looks plain: that is expected. Each recognised line becomes a paragraph. Rebuild headings and tables in Word.
- The PDF asks for a password: unlock it first with Unlock PDF (you need the password).
- The scan is longer than 50 pages: run it in batches, or split it with Split PDF.
Only need the text, not Word?
Use Extract PDF Text to copy the recognised text or save it as a .txt file. To translate the text into English, use Translate PDF.