OCR PDF
Turn a scanned PDF into one you can search, select and copy from. The pages look exactly the same; the words go in as an invisible layer on top. It all runs in your browser, so your scan never leaves this device.
Drop a scanned PDF here
Your file stays on this device. Nothing is uploaded.
How to OCR a scanned PDF and make it searchable
- Drop the scanned PDF onto the box above, or press “Choose a PDF file”. The page checks which pages are pictures and which already have text.
- Pick the language the document is written in. Its download size is shown before anything downloads.
- Press “Make searchable & download” and wait while each page is read. “Stop” cancels at any point.
- Open the downloaded PDF and search it. “Download the text (.txt)” gives you the words on their own.
What OCR PDF does
Scanners and phone scanning apps save each page as a picture. The PDF looks like a document and behaves like a photograph: you cannot search it, select a sentence, or have it read aloud. This reads the words in each picture and lays them over the page as invisible text, so the same file becomes searchable.
What it does
- Reads the printed text on scanned pages and adds it as an invisible, searchable layer, page by page
- Leaves every page looking exactly as it did; nothing is redrawn or re-compressed
- Skips the pages that already have text, and tells you which ones
- Reads sixteen languages, Chinese, Japanese, Korean, Hindi and Nepali among them, each downloaded from this site only when you pick it
- Gives you the recognised text as a .txt file as well
What it does not do
- Upload your scan; the reading engine runs in this browser tab
- Read handwriting
- Rescue a poor scan: faint, blurred, tiny or crooked print comes out with mistakes, and the pages it found hard are named
- Rebuild the layout as an editable document; PDF to Word can take the text out of the searchable file afterwards
- Read Arabic or other right-to-left scripts yet
- Turn sideways pages upright; Rotate PDF does that first
Last updated
Need to send this to someone?
Our sister site QRSansar hosts a document and gives you a short link and a QR code for it. That is handy for WhatsApp, a slide, a poster or a printed handout, and it avoids attachment size limits entirely. You can swap the file later and the link and QR code keep working.
Worth being clear about the difference: that is a hosting service, so your file is uploaded to it. Everything on PDFsansar stays on your device; sharing is the one job that cannot.
Share a document on QRSansar →Other tools
All free, and all run in this tab without uploading anything.
OCR PDF: common questions
Is my scan uploaded?
No. The reading engine, Tesseract, runs in this browser tab. The first time you use it, the engine and the language you pick download from this site, not from anybody else's, and your browser keeps them. The PDF itself is read and rewritten on this device and never sent anywhere.
Will the pages look any different?
No. Every page stays exactly as it was. The words go in as an invisible layer, placed over where they appear in the scan, so when you select a sentence the highlight lands on the printed words. Nothing is redrawn or re-compressed, and the file grows by a few kilobytes a page.
Which languages can it read?
English, Spanish, French, German, Portuguese, Italian, Indonesian, Malay, Filipino, Swahili, Vietnamese, Hindi, Nepali, Simplified Chinese, Japanese and Korean. Pick the one the document is written in; each downloads only when you choose it, and its size is shown first. Arabic is not offered yet: it can be read, but some PDF readers show the result backwards, and a search that cannot find the words is no use.
How long does it take?
A few seconds a page on a computer. Measured on a 4-core machine in Chrome, it averaged 3 to 9 seconds a page: a receipt at the quick end, a full page of small magazine print at the slow end, and the busiest single page took 11. A phone is slower. The first time, the reading engine and the language download before anything is read, about 4 MB for English, and after that your browser keeps them. You can leave it running in another tab, and Stop works at any point.
How accurate is it?
On clean printed pages, nearly every word: on two magazine and report pages from a standard OCR test set, 99% of the words came back. On real scanned forms in German, Spanish, French, Japanese and Chinese, 91 to 96%. On scanned shop receipts, small print on thermal paper, 79%. Mistakes gather in names, numbers and small print, so check those before you rely on them. The pages it found hard to read are named when it finishes.
What happens to pages that already have text?
They are left exactly as they are, and the page tells you which ones before it starts. Reading them again would lay a second copy of every word over the first. If every page already has text, it says so and does nothing.
Can it read handwriting?
No. Tesseract is built for printed text. Handwriting comes out as nonsense or nothing at all, and a page with nothing readable on it is named when the job finishes.
Can I get the text on its own?
Yes. When it finishes there is a button for the text as a .txt file, with each page in order. Pages that already had text are included as they were.