The latest Data Phoenix news tagged text-to-image.
A transformer maps coarse image structure in patches, and a second network uses that map to update pixels across the whole image.