@Szasz I recon that AI markers can be yet another tensor layer to the diffusion model. Explained simply: 3d texture = tensor layer textures = tensor layer game logic (even if hidden) = tensor layer Everything you can annotate or somehow infer from already existing maps can teach the model. So if you have good maps with already existing AI markers, the model can learn from it and create AI markers for new maps. Although I must say that is currently out of scope for my project