Skip to content
Esc

Arabic image to text (OCR)

Image never leaves this browser

Extract Arabic text from images, screenshots and scanned PDFs, returned right to left as Unicode you can paste anywhere. Persian and Urdu have their own models in the list.

For mixed text such as Hindi with English words
Helps photos of paper; turn off for coloured screenshots

How to use it

  1. Check that the language is Arabic; tick Also read English if the text mixes in English words.
  2. Drop the image or scanned PDF, or paste a screenshot with Ctrl+V. The first run downloads the language data.
  3. Copy the text or save it as a .txt file. Check names and numbers against the original.

Details

Engine
Tesseract 5 LSTM via tesseract.js 7 (Apache-2.0), running in a Web Worker in this tab
Language
Arabic, or any of 120+ Tesseract languages from the list. Mixed-language text: tick Also read English
Input
JPG, PNG, WebP, HEIC, screenshots pasted with Ctrl+V, and scanned PDFs (up to 30 pages, rendered at 200 DPI)
Download
The engine and the language data, a few megabytes each, fetched once from jsDelivr and cached by the browser
Output
Plain text you can copy or save as .txt, with Tesseract's confidence score per image

Questions

Is my image uploaded?

No. Text recognition runs in this tab with Tesseract compiled to WebAssembly. The only downloads are the engine and the language data (from the jsDelivr CDN, cached after the first time); your image is never part of any request.

Does it handle right-to-left text?

Yes. The text box switches to right-to-left for Arabic, Persian, Urdu and Hebrew, and the copied text keeps the correct order.

Are diacritics (tashkeel) kept?

Partly. Small vowel marks are often dropped or misread, especially in small print. Plain unvowelled text is read best.

Can it read Urdu Nastaliq?

Choose Urdu. Nastaliq calligraphy is harder for OCR than Naskh, so expect more errors with ornate fonts.