I have been wondering.. why not using just a toml print and parsing it with OCR?