Scanned PDF OCR
Recognize and pull text out of scanned/image PDFs that have no text layer · in your browser
1. Choose a PDF
2. Language
Language data downloads once on first use and is cached in your browser. Files are never sent to a server.
3. Run
Recognize and pull text out of scanned/image PDFs that have no text layer. Each page is rasterized and recognized with tesseract.js; files are never sent to a server.
What you can do
- ✓Extract text from scanned or photo PDFs
- ✓Korean+English, English, Japanese, Chinese recognition
- ✓View the result on screen, copy it, or save as TXT
- ✓Batch multiple pages (first 30)
- ✓in-browser — no upload
How to use
- 1Add a scanned PDF
- 2Pick the recognition language and click Recognize text
- 3Copy the result or download it as TXT
Supported formats · PDF (input) → text / TXT
Read text out of scanned PDFs whose pages are only pictures. Choose Korean+English, English, Japanese, or Simplified Chinese, then view the result on screen, copy it, or save it as TXT. Recognition runs inside your browser, so the document is never uploaded, and up to the first 30 pages are processed per run.
How to use it
- 1Select or drop the scanned PDF.
- 2Choose the language (Korean+English, English, Japanese, Chinese) and click Recognize. The first run downloads language data, so allow a moment.
- 3Review the recognized text on screen, then copy it or save as TXT.
When it comes in handy
Copying a passage from a scanned paper or court ruling, turning a contract or minutes that exist only on paper into searchable text, or moving numbers from a scanned table into Excel.
PDF to Word and PDF to Excel find nothing in a scanned PDF precisely because it has no text; this tool creates that text.
What makes recognition accurate
Printed text scanned crisply at around 300 DPI recognizes well. Blurry faxes, tilted phone captures, handwriting, and forms with patterned backgrounds increase errors. Matching the language to the document matters too; for Korean documents with some English, choose Korean+English.
Always compare the result against the original. Look-alike characters such as 0 and O, or 1 and l, are commonly swapped.
Scope and timing
Up to the first 30 pages are recognized per run. For longer documents, split them into 30-page chunks with Extract Pages. Recognition takes a few seconds to over ten per page depending on your device, and you can use other tabs meanwhile.
Language data is downloaded once and cached by the browser, so later runs start immediately.
Frequently asked questions
Are files uploaded to a server?+
No. Recognition happens inside your browser and the document never leaves your device.
How accurate is it?+
High for crisp printed scans; errors rise with blur, tilt, or handwriting. Compare the result with the original.
Which languages are supported?+
Korean+English, English, Japanese, and Simplified Chinese.
How many pages can it handle?+
The first 30 pages per run. Split longer documents with Extract Pages.
Can it produce a searchable PDF?+
This tool focuses on extracting text for copying and TXT export. You can paste the text into Word and save that as a PDF.
Related tools