same than #678 but for OCR

not familiar yet what are the best way to run your own OCR API, probably plenty of solutions

also maybe this could be a thing: https://github.com/microsoft/OmniParser

/bounty 150

please suggest which OCR engine would you run, how, what's the perf compared to apple native/windows native/tesseract/unstructured.io, etc, before doing the code