PDF OCR
Recognize text in scanned PDFs and export TXT, JSON, or searchable PDF
About this tool
PDF OCR turns scanned or image-only PDF pages into editable text. It renders the selected pages at a configurable scale, runs a real Tesseract OCR model in a browser worker, and reports page-level confidence so you know what deserves review. Results can be exported as plain TXT or structured JSON, and the optional searchable-PDF output combines OCR text with the rendered page images. The engine and language data are loaded only when you start recognition. This is different from a text-layer extractor: it can recognize pixels, but names, tables, dates, and low-quality scans still require human verification.
Scenario Recipes
Turn a scanned PDF into a searchable review copy
Goal: Recognize printed pages and produce text that can be searched and manually verified.
- Choose the page range, document language, and a higher render scale for small print.
- Run OCR, then review low-confidence pages, names, dates, and amounts in the extracted text.
- Export corrected TXT or JSON and create a searchable PDF only after the review.
Result: You get a searchable working copy plus editable text, with uncertain recognition checked before use.
Frequently Asked Questions
How is this different from extracting PDF text?
Text extraction reads an existing selectable text layer. OCR renders page pixels and recognizes them, so it can handle scans that have no text layer.
Does OCR upload my document?
The recognition job runs in a browser worker. The Tesseract engine and public language files are downloaded on demand, but your PDF pages are not sent to a ToolsKit server.
Can it create a searchable PDF?
Yes. Enable the option to combine the OCR result with rendered page images. Review the output because layout fidelity varies by scan.
Which languages are supported?
The interface currently offers English, Simplified Chinese, or a combined English plus Simplified Chinese model.
Why are tables or columns misread?
OCR recognizes text regions rather than spreadsheet structure. Increase render scale, use a suitable page layout, and verify the exported text manually.
Why does the first run take longer?
The browser must download and initialize the OCR worker, core, and selected language data. Subsequent runs can reuse cached files.
Keep browsing