My second model!

I used around 998 images to train this LoRA, so I imagine you can get quite a lot out of it (at least I hope so lol). Since most of the training images were monochromatic, I suggest adding the following tags to get better results, or even using it together with another LoRA that has flat colors.

Positive tags:

kon-kit, thicker outlines, flat colors, vibrant colors (maybe), high contrast (maybe)

Negative tags:

creature, monster, white pale skin, grey scale, monochrome,

During training I added tags referring to character names, but I must admit that I basically only tested it with two of them, and for one of them (makishima kurara ora clara sensei) I haven’t been able to get the results I wanted yet.

Here you can do whatever you want but my results were better with those values:

strenght_model: 0.4

strenght_clip: 0.8

...by the way, I use ComfyUI for generate those images.

Kon-kit Characters:

celes (holy angel)

black Knight sara

kaya izumi

yoriko tanaka

yuzuka okamoto

lamia

asahina meo

asahina mari (Mighty Wife)

Cherry Queen

Priestess (Three Heroes Adventures)

Koriyama Aiko (Proprietress)

Washiu Shirayuki

makishima kurara

kunoichi amame

kunoichi madoi

paco

hime (milday)

mira sensei

Mitaka Midori

sayuki

hoshino sensei

mitaka yukari