Arabic OCR — Extract Arabic Text From Any Image
Upload a photo, scan or screenshot containing Arabic and get back editable Arabic text in the original script. The model is told in advance that it is reading Arabic, so it keeps letters in their connected forms instead of transliterating them into Latin characters.
Drop your image here
or click to browse — JPG, PNG, WEBP, GIF…
Upload an image and click Extract Text
All readable text from the image will appear here.
What Makes Arabic OCR Difficult
Arabic is genuinely hard for OCR because each letter changes shape depending on its position in the word — isolated, initial, medial or final. The same letter ع has four distinct glyphs. Letters also join into ligatures (لا being the classic case), and short vowels appear as optional marks above and below the baseline. Generic OCR engines frequently drop those marks or split a single cursive word into several fragments. Reading right-to-left adds a second failure mode: mixed Arabic-and-numerals lines, like a date or an invoice total, often come out with the digit groups reordered.
Getting Accurate Arabic Results
- Photograph the page straight on. Because Arabic is cursive, even a 5° skew merges adjacent letters into shapes the model has never seen.
- Keep harakat (vowel marks) in frame if you need them — crop above the line and below the descenders, not tight to the letters.
- For Quranic or poetry text with full vocalisation, expect to proofread the marks; the consonantal skeleton is far more reliable than the diacritics.
- Naskh and Kufi print extract cleanly. Decorative Diwani or Thuluth calligraphy is much less reliable — treat the output as a draft.
Who Uses Arabic OCR
Students and researchers
Digitise pages from Arabic textbooks, manuscripts and printed journals so the text becomes searchable and quotable.
Business documents
Pull supplier names, totals and dates off Arabic invoices, contracts and official correspondence without retyping right-to-left text.
Social and messaging screenshots
Copy Arabic text out of WhatsApp screenshots, posts and images where the text is not selectable.
Arabic OCR — Frequently Asked Questions
Does the Arabic OCR keep the original script or transliterate it?▼
It keeps the original Arabic script. The extracted text comes back in Arabic characters, so you can paste it straight into a document, spreadsheet or translation tool. Nothing is romanised.
Can it read Arabic handwriting?▼
Clear, separated handwriting often works. Joined everyday handwriting and formal calligraphic styles are considerably less accurate because the letterforms deviate far from printed Naskh. Printed or typed Arabic gives the best results by a wide margin.
Are diacritics and hamza preserved?▼
Hamza and the dots that distinguish letters like ب ت ث are usually preserved reliably. Full vocalisation marks (fatha, kasra, damma, sukun) are extracted less consistently and should be proofread if they matter for your use case.
Will mixed Arabic and English text work?▼
Yes. Bilingual documents are common and the model handles both scripts in one pass. Check the ordering around numbers and Latin words, since right-to-left and left-to-right runs on the same line can occasionally come back in the wrong sequence.
Need a Translation Instead?
This page extracts Arabic text as written. If you want the meaning in another language, the image translator reads the text and translates it in one step.
OCR in Other Languages
Or use the general image to text converter if your image contains several languages at once.