Image to Text (OCR)

Copy the text out of a photo, screenshot or scan, in English or Arabic. The text reader runs on your device, so the image is never uploaded.

Loading tool…

How a picture of text becomes text

To a computer, a photo of a page is only coloured dots; the letters in it cannot be copied, searched or edited. OCR (optical character recognition) finds the lines of text in the picture, then reads each line as a sequence of letters. This tool uses Tesseract, the open-source OCR engine first developed at HP and later at Google, whose reader is a neural network trained on printed text in more than a hundred languages; here it reads English and Arabic.

Arabic is harder than English for OCR: letters join and change shape, many differ only by their dots (ب ت ث ن ي), and vowel marks sit above and below the line. In our own tests on printed text drawn in a common Arabic font, it got 98 to 99.5 of every 100 Arabic characters right (a character error rate of 0.5% to 2.3%, worst on a low-resolution image), and English 99.5 to 100 in 100 except on that low-resolution image (90 in 100). Those were clean, computer-made pictures. Phone photos of paper, uneven light, curved pages, decorative or calligraphic fonts and handwriting do noticeably worse, so proofread the result, above all names and numbers.

Why we built this tool

Use it when the words you need are stuck in a picture: a screenshot of a message, a photo of a notice or a menu, a scanned letter or certificate, a slide photographed in a lecture. Copy the text, or download it as a .txt file, then edit, translate or search it.

Many OCR sites upload your image to their servers, and the pictures people read this way are often private: ID cards, contracts, medical letters. Here the text reader is downloaded to your browser the first time you use it (about 3 MB for Arabic, 4.5 MB for English, 6 MB for both) and runs on your device, so the image is never uploaded. There is no account and no page limit.

Tips and common mistakes

  • Choose the language the text is in. “Arabic and English” is for pages that mix both; for a page in one language, choosing just that one is more accurate.
  • Photograph the page straight on, in even light, filling the frame. A sharp, level picture matters more than a high megapixel count.
  • Crop away everything that is not text (with Crop Image) before reading: pictures and borders can be read as stray letters.
  • Text smaller than about 10 pixels high is hard to read; take a closer photo instead of zooming in afterwards.
  • For a scanned PDF, use Extract Text from PDF: it reads pages that have a text layer directly, and offers this same OCR for the scanned ones.

AI on this page

  • Model: Tesseract OCR (Tesseract.js 5.1.1) with the tessdata_best Arabic and English language data
  • What it does here: Reads the letters in your image and turns them into text you can copy or save: it finds the lines of text, then recognises each word with a neural network (LSTM) trained for that language.
  • Download: About 3 MB for Arabic, 4.5 MB for English or 6 MB for both, engine included, downloaded once per language.
  • Licence: Apache 2.0 (engine and language data).
  • Privacy: Nothing is downloaded until you press Read text. The reader runs on your device; your images are not uploaded.
  • Limits and speed: On clear, computer-made text in our tests it got 98 to 99.5 in 100 Arabic characters and almost every English one right. Photos taken at an angle, low resolution, handwriting and calligraphic fonts do much worse. A picture took 0.3 to 2 seconds in Chrome and Safari, 4 to 6 seconds in Firefox. Causes and fixes for wrong text: Why does OCR get my text wrong?

Quick start: using Image to Text (OCR)

  1. Choose one or more images: photos, screenshots or scans.
  2. Pick the language of the text.
  3. Press Read text. The first time, the text reader downloads; after that it starts at once.
  4. Check the text, correct anything misread, then copy it or download it as .txt.

For a scanned PDF, use Extract Text from PDF; to straighten and trim a photo first, Crop Image and Rotate & Flip Image.

Frequently asked questions

Is my image uploaded?

No. The OCR engine (Tesseract.js) and its language data are downloaded to your browser and run there. The image is read from your device and never sent anywhere.

How accurate is the Arabic?

On clear printed text it is good: in our tests on computer-drawn Arabic text it got 98 to 99.5 characters in 100 right. On phone photos, decorative fonts and low-resolution images it makes more mistakes, typically confusing letters that differ by their dots. Always proofread.

Can it read handwriting?

Not reliably. Tesseract is trained on printed text; neat block capitals sometimes work, joined handwriting and handwritten Arabic mostly do not.

Why is the first run slow?

It downloads the reader, about 3 to 6 MB depending on the language, and the browser keeps it. Later runs start at once and take a second or a few per image (longer in Firefox and on older phones).

Can I get the text as a Word document?

Yes: “Download Word (.docx)” saves the text as paragraphs, with Arabic lines as right-to-left paragraphs and English words and numbers inside them kept in the right order. It opens in Word, LibreOffice and Google Docs. Only the text is kept: tables, columns, fonts and the layout of the picture are not rebuilt, so check and format it there.

Related Image Tools

Browse all Image Tools →