Handwritten & Scanned Document OCR
Turn handwritten and scanned pages into editable text, exactly as written.
Upload a photograph, scan, PDF or Word file and Qaaf AI reads what is written and returns it as editable text in the original language and script. Built for the documents ordinary OCR gives up on: handwritten Urdu and Arabic manuscripts, marginal notes, mixed scripts on the same page, and old madrasa records. Headings, lists, tables and poetry keep their structure, and anything genuinely unreadable is marked rather than guessed at.
Try it free — no account needed
100 free credits right now — no sign-up, no card. Upload a file and see the quality for yourself. You get 100 free credits right now, with no sign-up and no card. When they run out, a free account gives you 200 more.
Long documents are processed a few pages at a time, with a pause between sections. You can close this page — the work carries on, and the result waits in My Library.
A section could not be read after three attempts.
This usually means the scan is too blurred or too dark to read. Try a sharper photo, or send it to us at +92 330 2527151 and we will look at it.
Guides for this workflow
All guides →- How to Translate Arabic Fiqh Books to Pashto and Sindhi
- ترجمة الكتب الإسلامية: How to Translate Arabic PDF Books to English with AI
- The Publisher’s Guide: How to Translate Arabic PDF Books to English Accurately
Browse every language pair on the supported languages page.
Frequently asked questions
Can it read handwritten Arabic and Urdu manuscripts?
Yes, including Nastaliq and Naskh, marginalia and old scripts. Conventional OCR segments characters the way printed Latin type is segmented, which does not describe connected Nastaliq at all, so it returns fluent nonsense instead of an error. This tool reads the page in context instead.
What happens when a page is too damaged to read?
It is marked as unreadable rather than guessed. In religious material a plausible wrong word does more damage than a visible gap, because nobody catches it later. Faint scans usually work; barely legible pages can fail, and you are told which ones.
Does it keep the page numbers and order?
Yes. Pages are processed in order and the output keeps the original pagination, so a digitized manuscript can still be cited against the physical copy. Sections are never allowed to jump ahead of earlier ones in the queue.
Can it handle a whole archive rather than one book?
Yes, through the book and archive digitization service, which is the managed version of this tool. For a collection it is worth a free consultation on WhatsApp first, because the right approach depends on the condition and the script.
Is the extracted text editable?
Yes. The output is real text in a formatted document, not an image with a text layer glued behind it, so it can be searched, edited, typeset or translated afterwards. That is what makes the rest of the work possible at all.