Im starting a project that could need OCR and I'm not sure if doing it with open cv for preprocessing + tesseract, or use something like Google document AI which seems very good in a small test I run.