So I trained this Lora with Illustrious, but I generally run it on ComfyUI with my local graphics card and there it works pretty good. Do you have trouble with the model?

Stuhl tried it here on Illustrious checkpoint, got messy image

ashlander So I got some decent results using following specs, it still has its trouble with generating faces with larger characters, but I achieved better results with upscaling generated images, I‘ll link the second base model I used in the suggestions thingy next to WAI-NSFW (yk where you can suggest base models that work best with the lora). Here are the specs:

Positive Prompt: Dragontaur, large taur, large wings, large tail, muscular, long hair, beautiful face, good face, good quality, beautiful detailed eyes, masterpiece,best quality,amazing quality

Negative prompt: bad quality,worst quality,worst detail,sketch,censor,chibi,monochrome ,furry,colored sclera, hololive,extra eyes,multi breast,shoes,animal hands,((simple background)),

Steps: 40, CFG scale: 7, Sampler: DPM++ 2M Karras, Seed: 3064985113, Clip skip: 2, quantity: 2, draft: false, aspectRatio: 1:1, baseModel: Illustrious, width: 1024, height: 1024, workflow: txt2img, experimental: false, disablePoi: true