A model should take its training from public domain amd authorized parties not web scraping

At least you got part of your goals up its progress man