Skip to content

Tesseract

Tesseract reads printed text from image files using optical character recognition, or OCR.

Visit Tesseract

What does Tesseract need from a scanned PDF?

Tesseract reads images. If the source is a PDF made from scans, convert its pages to images first. Choose the appropriate language data and check a sample for missing words, mistaken characters and reading order.

What should your app keep besides the recognized text?

The engine is Apache 2.0-licensed and can run on your own computer or server. It produces outputs such as plain text, searchable PDF or TSV, a tab-separated table with text and position information. Keep the original scan, chosen language and output. If your search system creates vectors, that is a separate step.

Would you trust the first reading of a fictional scan?

Draw a low-quality inventory label with a code containing the number zero and the letter O.

  1. Compare the result with the image
    Mark which characters are easy to confuse and what the original says.
  2. Decide what can reach search
    Keep a reviewed text value with the source image reference.