Multimodal Generation
How does a sentence become an image? Not through one monolithic model — a text-to-image system is three separately-motivated trained components, wired together, and most of what a system can and can't do is a direct consequence of how they're joined.