update documentation

render all images ofline + update examples + compare with aws ocr

this has no changes to the functions, except the parts to use the "best" models