editing
about
This repository contains two Stable Diffusion LoRAs for generate "Cute MOE face" illustration, and their training data.
There was trained by kohya-ss's sd_scripts, and tested on AUTOMATIC1111/stable-diffusion-webui.
The base model is almost AOM3, the model I used was added a little merge as: AOM3.safetensors + Hipoly 3D Model LoRA v2.0:0.15:0,1,1,1,1,1,1,0.1,0.1,0.1,0.1,1,1,1,0.3,0.1,0.1. Hipoly 3d LoRA was merged by SuperMerger.
They don't require "trigger word", simply changes illustration's style when applied.
I recommend to set weight between 0.05 to 0.4, and use LoRA Block Weight extension.
facelora1
First LoRA is actually not made for this purpose. I attempted to make "Uchinoko"(means my daughter, favorite original character, lol) LoRA.
Train data are my favorite images(those are generated above model by me), and augmentation images made by controlnet's reference-only.
This LoRA worked well, worked better than I intended, and could be used to generate beautiful girls in a generic way.
facelora2
I generated various beautiful girls images with above LoRA, and made second LoRA trained by them.
This LoRA is more suitable for general purpose generation.