Make a Scanned PDF Searchable in Hindi or Tamil

What you get back, and what you do not

A scanned PDF holds each page as a picture. You can read it; Ctrl+F cannot. Make searchable renders every page at 300 dpi, recognises the printed words, and writes them back as an invisible layer underneath the scan — one text object per word, stretched to the box that word occupies on the page.

So search, copy and screen readers start working, and the page looks precisely as it did before, because the image was never touched. This is not a route to editing: a recognised word sitting under a scan is not text you can retype, and editing existing PDF text is a different job on a different kind of file. Handwriting is not supported at all.

Tick every script the page uses. Only those.

Forty-seven languages ship inside the package. The Indian ones are Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Nepali, Odia, Punjabi, Sinhala, Tamil, Telugu and Urdu, alongside English; the rest cover western Europe, Cyrillic and Greek, Chinese, Japanese, Korean, Arabic, Hebrew, Thai, Vietnamese, Indonesian, Malay and Swahili. Each one is a checkbox because the file is genuinely there beside the application; English is ticked when the panel opens.

A certificate with a Marathi heading and an English name needs both ticked. A page that is entirely Hindi wants Hindi and nothing else — a spare language does not cost you nothing, it costs accuracy, because the engine now has more ways to be wrong about the same smudge. Indic text also needs a face with the right glyphs, so those documents embed a subset of Nirmala UI, which is present on every supported version of Windows.

Run it, then check the words you actually care about

Start with a scan that is straight and legible; recognition cannot recover what the scanner never captured. Choose OCR — the panel is headed Make searchable — tick your languages and press Recognise. Leave Skip pages that already have text ticked when a document is part typed and part scanned. Nothing is written yet: the recognised layer goes into the edit session, the banner across the top says so, and the file on disk is untouched until you choose Save or Save a copy. On the command line it is the opposite - creasepoint ocr wants --out and writes a new file.

You get counts back: pages recognised, pages skipped, words found. Search the output for a phrase you can see on the page, then check the things that are expensive to get wrong — a registration number, a date, a surname with an unusual spelling — against the image. The image is the record; the layer under it is a machine’s best reading of it.

OCR, Word export, and a folder full of scans

Use OCR when you want a scan you can search and still hand over as the scan. Use Word export when you want the words out and into something you are writing — it recovers the text, its type and its pictures, and leaves the layout and the tables behind.

For a folder, Batch has an OCR job, but it is English only; anything in Marathi or Tamil goes through the interactive tool one document at a time, or the command line. All of it runs here. This application has no server, no account and nothing in it that sends a document anywhere. A separate cloud product is planned under the same name, and does not exist yet — if you ever use that one you are uploading on purpose, and this application is not involved.

Common questions