pdf2text and tesseract, i believe pdf2text uses tesseract. i have them installed on an apple silicon mac with homebrew (e.g. brew install tesseract or brew install pdf2text)

could probably use some ai computer vision package (i haven't checked, i remember looking around before settling on pdf2text) like opencv.

when i used pdf2text it was with pdf slides my prof provided, they ONLY gave pdfs. something about copyright and IP. super interesting prof, great scientist, great researcher, actually a member of some cool orgs like Linnaeus Society, and annoying with her lecture files.

EDIT: if anyone wants it enough i can try to do a proof-of-concept for like ~15 random pages of a random doc and see how well it goes