Reading a document photo and copying out the ID number, name, and date of birth by hand is tedious and error-prone. The new Indian Document OCR tool automates that first pass for five common Indian documents — Aadhaar Card, PAN Card, Driving Licence, Passport, and GST Registration Certificate.
How it actually works
The tool runs the same in-browser OCR engine as our Image to Text (OCR) tool — nothing is uploaded to a server. Once the raw text is recognized on your device, a second step runs entirely in your browser too: it pattern-matches that text against the format each document type is known to use (Aadhaar's 12-digit number, PAN's 10-character format, a passport's machine-readable zone, and so on) and surfaces the fields it finds.
What "best-effort" actually means here
This is text extraction, not identity verification. The tool has no connection to UIDAI, the Income Tax Department, DigiLocker, or any other government database — it cannot confirm a document is genuine, active, or actually yours. It also can't guarantee every field is found: a blurry photo, an unusual layout, or a name printed in a stylized font can all cause a field to be missed or misread. That's exactly why the full raw recognized text is always shown alongside the extracted fields, so you can check anything that matters against the original document before relying on it.
Passport extraction uses the MRZ when it can
For passports specifically, the tool first looks for the machine-readable zone — the two lines of <-padded text at the bottom of the photo page — since it's a far more reliable source for the passport number, date of birth, sex, and expiry date than the free-form text elsewhere on the page. If the MRZ isn't clearly recognized, it falls back to matching labeled fields in the visible text instead.
A privacy-sensitive tool, on purpose
Aadhaar, PAN, and passport numbers are exactly the kind of data you don't want passing through a third-party server, so this tool is designed the same way our AI-powered Background Remover and OCR tools already are — the recognition model runs on your own device, and the image, the recognized text, and the extracted fields never leave your browser.
Using the tool
Pick the document type, choose a language if the document mixes English with Hindi or another script, and upload a clear photo or scan. Try it here: Indian Document OCR.