Image to searchable PDF — free private OCR

Turn photos into a searchable PDF — free OCR that never uploads your files

Updated August 16, 2026

Drop in photos of documents and get back one PDF you can search, select, and copy from. The text recognition runs in your browser with the open-source Tesseract engine — your files are never sent to a server.

Runs entirely in your browser — your files are never uploaded anywhere.

Drop images here or tap to choose
JPG, PNG, or WebP — photos of documents work best. Add as many pages as you need.

Everything happens on your device. Your images are opened in this page's memory and never sent anywhere — the only network requests fetch the open-source OCR engine and, once per language, its recognition data. Large photos are downscaled to ~2,500 px for recognition; the PDF embeds a high-quality copy of the original.

How to turn photos into a searchable PDF

Add your images. Drop JPG, PNG, or WebP photos onto the tool, or tap to pick them. Each page appears as a card — drag the handle to reorder, or remove the ones you don't want.
Pick the language. Choose the language the documents are written in. The recognition data for that language downloads once (the size is shown next to the picker) — your images still never leave your device.
Recognize. Hit Recognize text and watch each page's card walk from waiting to done, with a live word count. You can cancel mid-run; nothing is kept.
Check and download. The recognized text is shown per page so you can verify it before trusting it. Name the output file, then download the searchable PDF, grab the plain .txt, or copy everything to the clipboard.

What it does

Real OCR in your browser — files are never uploaded anywhere
Multiple photos become one multi-page PDF, in the order you choose
Invisible text layer: the PDF looks like your photo but is searchable and copyable
15 recognition languages, downloaded on demand
Also exports the plain recognized text as a .txt file

What a searchable PDF actually is

A photo of a receipt is just pixels — you cannot search it, select the total, or copy the merchant's name. A searchable PDF adds a second, invisible layer on top of the image: the recognized text, positioned where it appears in the picture. The page still looks exactly like your photo, but Preview, Adobe Reader, Spotlight, and every document system can now read it.

This tool builds that file in your browser. It decodes your images, runs the open-source Tesseract recognition engine compiled to WebAssembly, and assembles a PDF with pdf-lib — all in the page you are looking at. Nothing is transmitted. That is the whole point: receipts, contracts, medical letters and ID documents are exactly the files you should not hand to a random converter website in exchange for free OCR.

When to use it

Photos of paper documents that you want findable later — receipts, warranties, letters, forms.
Scans that came out of a scanner as plain images.
Anything you would not paste into a stranger's website.

Why position matters in the text layer

The recognized words are not just appended to the file — each one is placed invisibly at the coordinates where it appears in the photo. That is what makes selection feel natural: drag across the image and the highlight follows the printed lines, search for a word and the viewer jumps to the right spot on the page. The geometry is approximate, the same way it is in files produced by office scanners, and that is fine for finding and copying.

Honest limitations

On-device OCR is slower than a server farm, and it reads print, not handwriting. Very small or blurry text may come out wrong — the .txt export makes it easy to check. Images are downscaled to about 2,500 px on the long edge for recognition speed; the embedded page image keeps up to 4,000 px, which prints cleanly at letter size.

If you do this often

This tool is a one-shot converter. If your camera roll is where documents go to disappear, that is the exact problem Paperlock exists for: it scans, reads, and files your paperwork on your iPhone — with the same rule that nothing leaves your device unencrypted.

Common questions

Is my document uploaded to a server?

No. Your images are opened in your browser's memory and processed there. The only network requests fetch the open-source OCR engine and the language data — they carry code to you, never your files. You can verify this in your browser's network inspector.

How accurate is the recognition?

The same as the Tesseract engine that powers many commercial scanners: very good on clearly printed text photographed straight-on, weaker on small print, skewed photos, and low light. Handwriting is mostly not readable. The tool shows you the recognized text per page — and flags pages that came back with low confidence — so check before relying on it.

Why is it slower than online OCR sites?

Online converters run on server hardware — and see your documents in the process. This tool trades some speed for privacy by doing the same work on your device. Photos are downscaled to about 2,500 pixels on the long edge before recognition to keep it reasonable; the PDF still embeds a high-quality copy of your original.

Can I feed it an existing scanned PDF?

Not yet — it currently accepts images (JPG, PNG, WebP). Export the pages as images first, or photograph the document. PDF input is on the list.

Which languages are supported?

English, Spanish, German, French, Portuguese, Italian, Dutch, Polish, Russian, Ukrainian, Turkish, Czech, Swedish, Danish, and Norwegian. Each language pack downloads once (a few megabytes) the first time you use it.

Does the searchable text look right when I select it?

The words sit invisibly on top of the photo at roughly the position they appear in the image, so selection and search highlight sensible regions. The geometry is approximate — that is normal for OCR text layers, including the ones professional scanners produce.