Bob Coyne has been working on a system for generating images of still scenes from text descriptions for about 15 years now:
https://www.wordseye.com/ http://www.cs.columbia.edu/~coyne/papers/wordseye_siggraph.p...