1. Home
  2. PDF
  3. OCR PDF

OCR PDF

Recognize text in scanned PDFs on your device and make them searchable.

Runs in your browser · No upload

How to use

  1. Drop a scanned PDF or a photo (JPG, PNG, WebP, BMP or TIFF). For a password-protected PDF, enter its open password.
  2. Choose the document language (Korean + English, Korean or English) and the pages to recognize. One run handles up to 200 pages.
  3. If the text is faint or small, raise the resolution (200, 300 or 400 dpi) or turn on preprocessing.
  4. Press Start recognition. Progress is shown page by page, and you can cancel at any time.
  5. When it finishes, download the searchable PDF and the TXT, or copy the recognized text.

Options

Option What it does
Resolution The size PDF pages are rendered at before recognition. 300 dpi is the default; use 400 dpi for small text or 200 dpi for speed.
Page layout Automatic (default) or tables / many paragraphs (assumes text flowing in a single column). Try both on table-heavy pages.
Convert to grayscale Drops colour and keeps brightness only.
Auto contrast Ignores the darkest 1% and brightest 1% and stretches the remaining brightness over 0–255. Suits faded scans.
Binarize (Otsu) Picks a brightness cut-off automatically: text turns black, background white.
Recognize pages that already have text When off, pages that already carry a text layer (10 or more non-space characters) are copied untouched.

How the result is built

  • Searchable PDF: each recognized page is stored again as an image (JPEG quality 85%) with the recognized text laid over it, invisible. No font file is embedded, so Korean and other scripts can be selected, searched and copied. After building, the tool re-reads the result PDF and reports whether the text really comes back out.
  • Image files with a side over 4,000 px are scaled down (aspect ratio kept) before recognition, and the page size in the PDF is set so the long side is A4 length (842 pt). TIFF works only if your browser can open it, and only the first page is recognized.
  • Preprocessing is used for recognition only; the pictures in the result PDF stay as they were. There is no deskew yet.

Otsu cut-off example

For an image that is half brightness 60 (text) and half brightness 200 (background), the Otsu cut-off is 129, so 60 becomes black and 200 becomes white. Photos with shadows or stains cannot be split by one cut-off and strokes can break; in that case turn binarize off and use auto contrast alone.

Accuracy and speed

Confidence Label Meaning
80% or more High Usually reliable. Still double-check numbers and proper names.
60–79% Medium Expect occasional mistakes.
Below 60% Low Raise the resolution or turn preprocessing on.
  • Printed text reads fairly well; handwriting, decorative fonts and very small text do not. Tables, vertical writing and mixed multi-column layouts are weaker, so check the result by eye.
  • Expect a few seconds to a dozen or so seconds per page, depending on your device and the resolution. Higher resolutions also use more memory.
  • On first use the OCR engine (about 4 MB) and the language data (about 3 MB for Korean + English) are downloaded once from this site, then reused from your browser’s storage.

FAQ

Does it work on PDFs that already have text? Pages with text are skipped and copied as they are. Turn the option on to recognize every page.

I only need the text. If the PDF has a text layer, PDF to Text is faster. This tool is for scans and photos with no text layer.

Related tools | Scan to PDF to bundle photos into a PDF, Compress PDF to shrink it, PDF to JPG to save pages as pictures, and Unlock PDF to remove a password.

Your files are processed only in your browser and are never sent to a server.

Can't find the tool you need?

Tell us what you want to do. We build the most requested tools first.