A model that generates outputs by taking environmental context and scene geometry as input conditioning.