How to extract Arabic text from an image or scanned file (OCR) with better accuracy
Last updated: November 17, 2026
Whenever you have a scanned document or a photo of a page and need to copy its text, you need OCR (optical character recognition). The result depends a lot on image quality, and this guide gives you the steps and tips for a better outcome.
Steps
- Open Extract text (OCR).
- Choose the image or scanned PDF.
- Select the language (Arabic, English or both).
- Start processing, then copy or download the text.
The tool runs on your device, so IDs and contracts are not sent to any server.
How to improve accuracy
Lighting and clarity. Photograph the page in even light without shadows, with the camera parallel to the page rather than tilted.
Resolution. A sharp image where the text is a readable size gives a better result. If the scan is low resolution, rescan at a higher one (about 300 dpi).
Orientation. If the page is upside down or tilted, rotate the image first.
Language. Pick the right language. If the document mixes Arabic and English, choose both.
Simple background. Black text on white paper works better than text over a picture or coloured background.
Review the result
The result is not always perfect, especially in Arabic:
- Numbers and dates: check them by hand, since they are the most common errors.
- Dots and diacritics: dots can be lost and similar letters confused.
- Tables: they sometimes come out as continuous text and need rearranging.
- Decorative fonts and handwriting: accuracy is lower.
So do not rely on extracted text in an official document without checking it against the original.
Turning a scanned PDF into text
If your PDF comes from a scanner, it is made of images. Turn its pages into images with PDF to images if needed, or feed the file straight to the tool.
Examples of suitable documents
- Suitable: a printed invoice, a book page in a clear font, a printed certificate, a screenshot of text.
- Less accurate: handwriting, decorative scripts (Diwani and Ruq’ah), tilted or shadowed photos, wrinkled paper.
Frequently asked questions
Does the tool support Arabic-Indic (١٢٣) and Western (123) digits? It usually recognizes both, but check digits manually because they are the most error-prone.
Why do strange characters appear between words? Usually low resolution, a busy background, or a logo or stamp over the text. Try cropping the part you need or re-taking the photo.
Can I extract text from a whole scanned PDF? Yes, but large files take longer on your device. Try a few pages first.
Is the diacritical marking (tashkeel) extracted? Usually it is missing or incomplete. For vocalized text, check the original.
How long does it take? It depends on the image size and your device’s speed. The first use may need to download the language model, which is then saved for later use.
Is it suitable for copying text from an ID or contract? Yes, and processing is local so your ID is not uploaded anywhere. But check the result against the original before any official use.