update SDXL version

I must say SDXL have far better results.

The lora have as flexible as the SD1.5 version while more details are involved and more precisely generated character.It can generate her ring and pants in some time if the prompts are right. You can see the examples for those prompt. But checkpoint models have great effect on this. Most pictures are generated in Yamersanime and KokakuXL

This model is hard to control due to the lack of the resources. Need to use lower weight, blockweight and/or strong prompt to do whatever you want. The blue jacket is hard to remove. It cannot generate the right pants as the original character should be. Accessories have too many details to regenerate.

So why did I upload this lora if it has so many shortcomings? Because after hours of trying, more than twenty versions are made, still cannot find a version which can generate better picture than this, not even a close one. I don't even know how this one is produced. Maybe one day I will upload a better one. But I have no clue now.

Easy to produce nsfw, but hard to hide the skin.