PDF OCR
Extract text from a scanned PDF page with the Tesseract OCR engine, running in your browser. Pick a page, run recognition, and copy the result.
Input
Drop a scanned PDF here
or click to browse
Settings
The AI OCR model downloads to your browser the first time it runs and is cached for future use. Works best on clean printed text.
How to OCR a scanned PDF
Pick a page, run OCR in your browser, and copy the recognised text.
- 01
Load the scanned PDF
The tool reads it in your browser and shows how many pages it has. If the PDF already has selectable text, PDF to Text is faster and exact.
- 02
Pick the page to read
OCR runs one page at a time. Enter the page number, counted from 1.
- 03
Run OCR and watch the progress
The page is rendered as an image at three times its normal size, then read by the Tesseract OCR engine. The first run downloads the engine and English language data, so it's slower; later pages are quicker.
- 04
Check and copy the text
Proofread numbers and names against the page preview before you use the text anywhere important.
How scanned pages are read
A scanned PDF is a set of photos of pages, with no text inside it. The tool draws the chosen page at 3× scale, which is about 216 dots per inch, so letters are large enough to recognise. It then runs Tesseract, an open-source OCR engine, compiled to WebAssembly so it works in your browser. Tesseract finds lines of text, splits them into words and recognises each one with an English language model.
Your PDF never leaves your device. The only downloads are the OCR engine and its English data, fetched once and then cached by your browser.
Which PDFs work well
The quality of the scan decides the quality of the text far more than anything else:
| Kind of page | Typical result | Tip |
|---|---|---|
| Clean office scan of typed text | Very good | Scan at 300 DPI if you control the scanner |
| Phone scan app output (flat, cropped) | Good | Use the app's document mode, not a plain photo |
| Faint photocopy or fax | Fair | Expect errors in small print and numbers |
| Two columns or tables | Text is right, layout isn't | Columns may run together; tidy by hand |
| Handwriting | Poor | Tesseract is trained on printed text |
| PDF that already has selectable text | Unnecessary | Use PDF to Text for an exact copy |
When OCR helps
Copying text from a scanned letter
Pull a paragraph out of a scanned notice or contract so you can quote it or translate it.
Searching an old scanned document
Run OCR on the pages you need and search the text, rather than reading every page by eye.
Retyping forms less
Extract the text from a scanned form, then correct the few misread words instead of typing it all again.
Limits worth knowing
- English only. Other languages and scripts aren't recognised.
- One page per run. For a long document, process the pages you actually need.
- The output is plain text. It doesn't add a searchable text layer to the PDF itself.
- Big pages at 3× scale use a lot of memory, so older phones may be slow or fail on dense pages.