OCR
When Han AI receives a scanned image or photo containing text, it extracts the words using Tesseract. Both English and Khmer language packs ship by default.
What it does
Converts an image into readable text on the VPS, with no upload to an external vision API.
| Field | Value |
|---|---|
| Schema name | ocr_image |
| Powered by | Tesseract OCR |
| Languages | English (eng), Khmer (khm) |
| API key required | No |
When Han AI uses it
- A scanned contract or receipt arrives as a photo or image PDF.
- A signage shot, business card, or printed menu needs to be read.
- A document extract returned empty text because the PDF has no OCR layer.
Examples
- “Read the text on this receipt and add it to expenses.”
- “What does the Khmer text on this notice say?”
- “Pull the supplier name and phone number from this business card photo.”
Limits
- Handwriting recognition is unreliable. Tesseract is built for printed text.
- Heavy skew, low resolution, or poor lighting degrade accuracy.
- Languages outside English and Khmer require adding their language pack —
TODO: confirm process for adding additional Tesseract language packs per tenant.
Why this stack
Tesseract is the industry-standard open-source OCR engine. Free, local, no third party sees your scanned documents.