LithoGAN reframes lithography simulation as an image-translation task, mapping mask layouts directly to resist patterns. Its reported results show substantial runtime reductions against conventional and machine-learning-based flows, while highlighting the importance of predicting both pattern shape and location.