I imagine all vision models have to do this, this being html rendering, to be able to do well in web design.