>>2874728I try to combine the known tags at high strength to create said outfits. Sometimes you get some unique outfits by doing so.
>>2874730Resolution sizes are based on what the models are trained around. SD1.5 models are usually 512x512/768. XL/Pony models are 1024/1024. If you stray away from that too much you start having bizarre deformities. This is why you rely on hi-res fix or tiled diffusion to upscale your gens.