PDF OCR: Extract Text from Scanned PDFs and Images

Recognise text in scanned PDFs and in JPG, PNG or WEBP images. Each page is rendered to an image and read by Tesseract.js in the language you pick. The result is plain text you can copy or save as a .txt file.

Drop your scanned PDF or image here

Supports PDF (up to 20 pages), JPG, PNG, WEBP - up to 50 MB

For regular PDFs with selectable text, use the PDF to Text Extractor instead

Runs in your browser: your files and text are not uploaded. Free, no account needed.

Last updated: · Reviewed by the QuickTooly Team

PDF OCR Guide

OCR or plain text extraction?

If you can highlight individual words in a PDF viewer, the file already contains real text. The PDF to Text extractor reads that text exactly, with no recognition errors, and handles any page count. Use OCR only when the page is a picture: scans, phone photos of paper, faxes, or PDFs where text selection grabs the whole page as one block.

How this tool reads a page

  • PDF pages are rendered with PDF.js at 2× scale, which is about 144 pixels per inch of page. Images are read at their own pixel size.
  • Recognition uses Tesseract.js version 5, a WebAssembly build of the open-source Tesseract OCR engine, running its LSTM neural-network recogniser.
  • The engine code is loaded from the jsDelivr CDN, and the trained language file is fetched from tessdata.projectnaptha.com the first time you run a language. Your document itself is not sent to either server.
  • Pages in the output are separated by a --- Page break --- line. The download is named after your file with _ocr.txt on the end.

Languages

The OCR Language menu offers English (default), Spanish, French, German, Italian and Portuguese. Only one language model is loaded per run. Each model is trained on its own language's alphabet and vocabulary, so accented letters such as é, ñ or ß are read more reliably with the matching model. For a page that mixes two of these languages, choose the one with more text.

Getting cleaner results

  • Straight, evenly lit pages work best. Rotate sideways scans upright first, and crop away dark borders from phone photos.
  • Very small print in a PDF may fall below what 144 pixels per inch can resolve. For that, export the page as a high-resolution image (3×) and run OCR on the image instead.
  • Tables, columns and handwriting are weak spots. Expect column text to be merged and table cells to lose their layout. Proofread names, numbers and amounts.

Limits

  • Files up to 50 MB; PDF, JPG, PNG and WEBP only.
  • Only the first 20 pages of a PDF are processed. Split longer files and run each part separately.
  • The output is plain text. The tool does not create a searchable PDF.

Frequently Asked Questions

What happens to pages after page 20?

They are skipped. Use the Split PDF tool to break the document into parts of 20 pages or fewer, then run OCR on each part.

Why is the first run slower than the next ones?

The first run downloads the Tesseract engine and the trained data for the chosen language. Your browser can cache these files, so later runs in the same language usually start sooner.

Can it read handwriting?

Tesseract is trained on printed text. Neat block capitals may partly work, but cursive handwriting usually comes out as noise.

My language is not in the list. What can I do?

Only the six listed languages are available here. Running OCR with the wrong language model produces many errors, especially on accented or non-Latin characters.

Related Tools

  • PDF to Text: exact extraction when the PDF already has a text layer.
  • Split PDF: break a long scan into parts under the 20-page limit.
  • Rotate PDF: turn sideways scans upright before recognition.
  • PDF to PNG: export a page at 3× to OCR very small print.