Practical tips for getting accurate Urdu text out of photos, screenshots and scanned pages with free OCR, and how to fix common mistakes afterwards.
Urdu is one of the harder scripts for a computer to read. The letters connect, they change shape depending on where they appear in a word, and small dots above and below a letter decide which letter it is. Even so, you can get useful results from a free in-browser OCR tool if you give it a good picture and know what to check afterwards. This guide covers both.
The result is that you should expect good but not perfect output on a clean printed page, and a lot of cleaning on a poor photo.
The quality of the input decides most of the output. Before you upload:
The guide to scanning with a phone camera covers these habits in more detail.
An OCR run uses one language setting. If a page is mostly Urdu with a few English words, choose Urdu and fix the English words by hand. If it is half and half, such as a form, run it twice: once on a picture cropped to the Urdu parts and once on the English parts. Numbers can come out in either Eastern Arabic or Western digits, so check phone numbers and dates carefully.
After extraction, read the text against the original and look especially for:
Paste the result into Word, Google Docs or a messaging app. Set the paragraph direction to right to left so punctuation lands on the correct side. To display Urdu in the familiar Nastaliq style, choose a Nastaliq font such as Jameel Noori Nastaleeq in your document; the OCR output itself is plain text and has no font.
If the Urdu is in a PDF and copying it gives broken text, the problem is probably the way the PDF stores its fonts, not the language. Read why you cannot copy text from a PDF, then either run OCR directly in PDF to Text or convert the pages to images and use Image to Text.
For a clean, printed, single-column page in a plain font, Urdu OCR can save you a lot of typing. For decorative calligraphy, old newsprint, faint photocopies and handwriting, treat the output as a rough first draft and be ready to type parts yourself. Your picture is processed in your browser and is not sent to any server.
Yes. Image to Text includes Urdu as a recognition language. Accuracy depends heavily on the print style and the quality of the picture, and printed text is much easier than handwriting.
Nastaliq letters slope diagonally, overlap and change shape with their neighbours, so a line has no straight baseline. Many OCR engines find this harder than the flatter Naskh style often used on screens and in newspapers.
Not reliably. OCR is designed for printed text. Handwriting varies from person to person, and the result usually needs heavy correction.
Pick the language that makes up most of the text. For better results, crop the English and Urdu parts into separate pictures and run each one with its own language.