Here we are again.. getting weird ;)

I trained this in a TOTALLY different way under the guidance of some pros, (thank you btw), so it may be a little different than v1 but in a better way.

The Pros: Adheres to prompts a lot better than my other loras. More or less, if it's organic or semi-organic, it should accept the lora's style. Works very well with other loras also and in my testing even works on pony models (tho I do not promise it works on all pony models).

The Cons: If you use a high CFG you will get overblown images. It can vary between models but I found between CFG 4-5.5 is a good range to prevent that.

Suggested weight is between 0.4-0.6 (on Forge/A1111, Comfy may need or less), when you get around the 0.8 weight of things you may notice some degradation if you are making something with a ton of small details, like scales but, in general, a hi-res fix, well, fixes it. (go figure, right?)

ALL EXAMPLE PHOTOS were made without hi-res fix!