You should just have it OCR a screenshot of the PDF that would probably work better