What you can do today

  • Recognize printed English on PDF pages without native text.
  • Keep word coordinates and recognition scores for source highlights and review.
  • Flag low-confidence words and affected invoice fields while retaining the original text.
  • Surface unreadable pages and OCR warnings instead of silently filling missing data.

A scanned supplier invoice

Upload a PDF scan, inspect the supplier, date, amount and line-item evidence, then correct uncertain values before export.

Supported scope and limits

  • PDF uploads only; standalone JPG, PNG and TIFF uploads are not supported.
  • Recognition currently uses English. Handwriting and multilingual recognition are not supported.
  • Pages with an existing text layer do not automatically get a second OCR pass. Poor scans need careful review.

Planned extensions

  • Language selection and mixed-language detection.
  • Image upload, deskew, orientation and image-quality checks.
  • Weak-text detection and selective reprocessing.

How to check the result

  • Read the original scan at a usable zoom.
  • Check digits, decimal points, dates and currency symbols.
  • Confirm all pages were processed; missing text is not a zero.