You photographed a page from a book, received a screenshot of a letter, or have a scanned PDF you need to edit. Retyping it all is slow. Optical Character Recognition (OCR) reads the letters in the image and turns them into real text.
How OCR works in PDFly
PDFly uses Tesseract, a respected open-source OCR engine, running entirely in your browser. You pick the language of the document, the engine analyses the shapes of the characters, and the recognised text appears in an editable box. It supports Arabic, French and English, as well as mixed documents such as Arabic + French.
Step by step
- Open Image to Text (OCR).
- Add a photo, screenshot or scanned PDF (up to 20 pages).
- Choose the language. For a French administrative letter with an Arabic header, choose Arabic + French.
- Press Recognise text. The first time, the language data is downloaded (a few MB); later runs are faster.
- Correct any mistakes, then copy the text or download it as a .txt file.
How to get the best accuracy
- Light: daylight or a bright lamp, no shadow from your phone.
- Angle: hold the phone parallel to the page so lines are straight.
- Focus: tap on the text before taking the picture.
- Resolution: the text should be at least 20 pixels tall. Zoom in rather than photographing a whole desk.
- Contrast: black text on white paper works best; coloured backgrounds reduce accuracy.
What OCR struggles with
Handwriting, stylised fonts, very small print, and complex tables are harder. Arabic with diacritics (tashkeel) or decorative calligraphy is less accurate than plain printed Arabic. Always proofread important text such as names and numbers.
OCR or PDF to Text?
If you can select text in your PDF viewer, the PDF already contains text: use PDF to Text, which is instant and 100% accurate. If you can't select anything, the PDF is a picture and needs OCR.