OCR PDF
Recognize the text in scanned PDFs and turn them into searchable, copyable PDFs and TXT files. Supports English and Korean, and your file is never uploaded.
Your files never leave this device. Everything is processed inside your browser and nothing is uploaded to a server.
How to use the OCR PDF tool
- Click "Choose PDF file" or drag your scanned PDF onto the screen.
- Choose the languages to recognize (English, Korean).
- The first time, click "Download data" to get the text recognition data. You will see the size before downloading.
- Click "Recognize text" and wait. Progress is shown page by page.
- When it is done, click "Download PDF" for a searchable, copyable PDF, or "Download TXT" for the text alone.
Frequently asked questions
Are my files uploaded to a server?
No. Files are processed only inside your browser and are never sent to any server. Only the code and character mapping data needed for processing are downloaded from this site; your files themselves never leave your device. You can check this yourself in the Network tab of your browser's developer tools.
Are my files safe even with ads on the page?
Yes. Ads only run outside the tool area. The tool area, where you choose and process files, is an isolated space your browser keeps separate, so other scripts on the page, including ads, cannot read your file content or file names. Nothing inside the tool area can be sent out. When you drag a file onto the page, the tool area expands to fill the screen so the isolated space receives the file.
What is OCR?
In a scanned or photographed PDF, each page is a picture, so you cannot search or copy the text. OCR (optical character recognition) reads the text in the picture and adds invisible text on top of the original page. The page looks the same, but you can search, copy and use read-aloud.
Why do I need to download data?
Recognizing text needs an engine and language data (about 4 MB for English, 7 MB with Korean). These are public files from this site. Once downloaded they are kept in this browser, so they are ready next time. Other browsers, devices and private windows need to download them again, and clearing browser data removes them. If the stored copy is removed, we ask again instead of downloading silently.
How accurate is it?
Clean scans of printed text are mostly read correctly. Blurry or skewed scans, handwriting, very small text and areas covered by tables or stamps may have mistakes, so check important content. The original page image is not changed, so mistakes are not visible on the page.
It takes a long time.
Each page is drawn at high resolution and read character by character, so it takes a few seconds per page. Several pages are read at once, depending on your CPU. Phones are slower and may stop if they run out of memory, so use a computer for long documents.
What does "Skip pages that already have text" do?
Pages whose body text can already be selected, such as PDFs saved from Word, are not read again. Scanned pages with only a page number or watermark as text are still read. This saves time on PDFs that mix scanned and digital pages, and avoids adding the text twice.
Is my file really not uploaded?
Yes. Only the recognition engine and language data are downloaded. Your PDF and the recognized text stay on this device. The recognition engine is Tesseract (Apache License 2.0).