@darius at any rate if you have a client that reformats something a miscellaneous user can already see these kinds of assists become viable. Run OCR on an uncaptioned image, etc.