Urdu OCR / PDF to Word

Read Urdu (and English) from a photo or scanned PDF on your device, then copy or download

Private mode processes files on your device. Optional High accuracy (AI) uploads one compressed page image at a time to our server and Google Gemini.

translate

Drop an image or scanned PDF, or click to upload

JPG, PNG, WEBP or PDF · max 30 MB · max 15 pages

Text on the page

Printed Naskh and PDFs with usable Urdu text usually read well. Always check names, numbers and reading order. Nastaliq may need corrections; High accuracy (AI) can help. Handwriting is not supported. For English-only scans, OCR Image to Text is lighter.

Jump to the next useful tool without leaving your flow.

See more

About this tool

Urdu is printed in two styles, and they behave very differently in OCR. Naskh (the style of most Word and web Urdu) reads almost perfectly. Nastaliq (books, newspapers, InPage and most printed notices) slopes, stacks letters and joins them into ligatures, which is why many free Urdu OCR tools return gibberish on it. This tool gives you two ways to read it. Private mode runs Tesseract on your own device with an Urdu model we trained on Nastaliq print and reads the page line by line - it never uploads, is exact on Naskh and good on clean Nastaliq print. High accuracy (AI) mode sends each page image to our server and Google Gemini, which reads Nastaliq far more reliably; nothing is stored, and it is clearly marked. PDFs that already contain real Urdu text are read directly, with no OCR at all. You can correct the result in place, then copy it, save .txt or download a proper right-to-left Word (.docx) file.

How to use this tool

  1. Drop an image or PDFJPG, PNG, WEBP or PDF, up to 30 MB and 15 pages. Larger page images are resized for safe processing; images above 40 megapixels must be resized first.
  2. Choose the language on the pageUrdu only is the most accurate setting. Pick Urdu + English only when the page really mixes both.
  3. Extract privately, or use High accuracy (AI)Private mode runs on your device. For Nastaliq books and newspapers, High accuracy (AI) reads far more exactly.
  4. Correct, then copy or downloadFix anything in the text box, then copy, save .txt or download a right-to-left Word (.docx) file.

How accurate is it?

We measured the tool on printed pages with known text (character error rate - the share of letters that come out wrong). Printed Naskh: about 0% in private mode. Printed Nastaliq: about 5-10% in private mode depending on the font, down from about 24% before the Nastaliq-trained model (October 2026); most remaining errors are dots and small marks, so plan to proofread. These figures come from clean printed pages in common Nastaliq fonts - phone photos, old scans and other fonts can come out worse. High accuracy (AI) mode is the one to use when every word matters - books, newspapers, legal pages and InPage output.

PDFs exported from Word or a website often carry the Urdu text itself. When that text is clean, the tool uses it directly and the result is exact. Many Urdu PDFs store broken glyph codes instead; the tool detects those and falls back to OCR automatically.

Private mode vs High accuracy (AI)

  • Private mode: runs entirely in your browser, nothing uploaded, works offline once loaded. Best for Naskh, clean print, and confidential documents.
  • High accuracy (AI): each page image is sent to our server and Google Gemini to be read, then discarded - we do not store it. Best for Nastaliq books, newspapers and InPage pages. A daily page allowance applies.
  • Both modes output standard Unicode Urdu (ی ک ہ, not the Arabic look-alikes), so the text searches, spell-checks and pastes into Word and InPage correctly.

Key features

  • Reads printed Naskh and usable PDF text layers; always check names, numbers and reading order.
  • Private on-device mode for confidential papers - nothing uploaded.
  • Optional AI mode for Nastaliq that ordinary OCR gets wrong.
  • Standard Unicode Urdu output, plus a real right-to-left .docx file.
  • Scanned PDFs as well as photos.

Common uses

  • Turning a printed Urdu letter, notice or circular into editable text.
  • Converting an Urdu PDF to Word for editing.
  • Reading a bilingual English/Urdu form.
  • Typing up a page from an Urdu book or newspaper with the AI mode.

Tips for better results

  • Choose Urdu only unless the page genuinely mixes in English - mixing the two lowers accuracy.
  • Shoot straight-on, well lit, one page filling the frame; crop away stamps and borders.
  • Use High accuracy (AI) for Nastaliq when every word matters; private mode for confidential pages.
  • Always check names, numbers and amounts against the original.

Mistakes to avoid

  • Expecting clean output from handwriting - both modes are built for print.
  • Leaving Urdu + English on for an all-Urdu page.
  • Sending OCR output to a court or office without proofreading it.
  • Using a tiny, compressed WhatsApp photo as the master scan.

Frequently asked questions

Yes. Private mode reads about 90-95% of the letters on a clean printed Nastaliq page in our tests and still needs proofreading. High accuracy (AI) mode reads Nastaliq much more reliably and is the right choice for books, newspapers and InPage pages.

Not in private mode - recognition runs on your device. High accuracy (AI) mode sends each page image to our server and Google Gemini to be read; the image is not stored. The button and the note under it say which mode you are using.

Yes. If the PDF contains real Urdu text, it is read directly and exactly; scanned PDFs are run through OCR. Download the result as a right-to-left Word (.docx) file. Layout is not copied - you get clean editable text.

Many Urdu PDFs store shaped glyph codes in visual order rather than real letters, so copy-paste produces nonsense. This tool checks for that and runs OCR instead when the stored text is broken.

No. Both modes are built for printed text. Handwritten Urdu gives poor results.

Yes, with no account. Private mode is unlimited; High accuracy (AI) mode has a daily page allowance.