>>70221680Thanks for the list.
>cosplay lora for 1.5HLL already has a lot of photos in its dataset, but if you try to prompt anything realistic it looks absolutely cursed. I don't think a separate cosplay lora will do it better.
I'll probably make one just to try out the new DoRA, but i doubt it will be good.
Have you tried to merge Animagine and something like Proteus, or using an anime model + "photo style lora". SDXL should be much better for this
