Did you go the “display it, then OCR what’s displayed” route as a last ditch effort?