Scans and photos

STYD reads scanned PDFs and photos of printed pages with text recognition (OCR) on its own server. The recognised text is then used like any other text from your files.

What is read

  • Scanned pages inside a PDF. A PDF that mixes typed pages and scanned pages keeps its typed pages as they are and reads only the scanned ones.
  • Photos (JPG and PNG, up to 50 megapixels). A photo is read like one scanned page and counts as one page.
  • Printed text in seven languages: English, French, Spanish, German, Italian, Portuguese and Thai.

STYD picks the language from the text already in the file, from your earlier files and from your interface language, and tries another one when the first result looks like noise. Tilted pages are straightened up to about 15 degrees, and upside-down pages are turned.

Where it runs

The recognition runs on STYD's own server, in a closed process with no network access. The page images are never sent to another company. Only the recognised text goes on to the next steps, which are described on What leaves STYD. Once the file is read, the original is deleted. The text STYD recognised is kept for search, like the text of any other file, until you delete the file.

Limits you should know

Recognition is not perfect. Always check a figure that comes from a scan against the page.
  • Printed text only. Handwriting is not supported. Stamps, signatures and logos are ignored.
  • Thai has more errors. Thai is read, but a few characters in every hundred can be wrong, for example lost tone marks and mixed-up Thai digits.
  • Tables are flattened. The text of a table in a scan is read line by line in reading order. The grid is not rebuilt, so columns can run together.
  • Unreadable pages are skipped. A page that is too faint, too small or too noisy is not added. The file's details list the pages STYD could not read. If no page can be read, the file fails.
  • Pictures inside Word and Excel files are not read. Only PDF pages and photos are.
  • Other languages are not read. Chinese, Japanese, Arabic, Russian and others work when the text is typed in the file, but not when it is a scan.

While it is running

A scan takes longer than a typed file: the file shows “Reading scanned pages: 3 of 12” while it works, and a large scan can take several minutes. Scans wait in their own queue, so ordinary files do not wait behind them. When the file is Ready, its details show how many pages were recognised, how many were hard to read and how many could not be read.

Tips for better results

  • Scan at 300 DPI, upright, dark text on a light page.
  • Photograph a page flat, in even light, filling the frame.
  • If a result looks wrong, add the file again after a better scan. The new file replaces the old one.