Generate the OCR on the fly, in the browser, when you do not have the proper OCR info. As someone that works on public web libraries, I see it useful (but wasteful)