Stable Diffusion 2.1 model trained, via Dreambooth, for generations 1-8 of Pokemon 3D (low poly..) renders. 768x768 for best results. Current state is inconsistent compared to 2D "Sugimori" model -- prompt detail is important. Upscaling also problematic (see 5-legged jolteon) Good for mixing species and types, adequate at applying species and types to non-pokemon terms (like fire horse).

Added terms include GMAX, Mega, (PokemonName), (PokemonType). GMAX works well as a style. Recommend adding "mega" and "gmax" to negative prompt unless you're into that.

"mega dragontype jolteon, claws, concept art" "firetype horse nidoqueen, horns" "dragon armored eevee, red, ((wings)), claws"