SEGA: Instructing Text-to-Image Models using Semantic Guidance Manuel Brack 1, 2 Felix Friedrich