lora anime

Table of Contents

Overview

Inspired by amazing work done by Trauter I decided to make a contribution to society by extending his work and developing new LoRAs. I'm going to train and test models on anime checkpoints like WarriorMama777, Andite, Gsdf, for that reason alone, I don't know how they will perform on your specific model.
You can find comparision grid in [model_name]/Previews folder.
Previews have metadata containing the prompt and settings used to create them, you can access this via "PNG Info" tab in Automatic1111/WebUI
Every model is trained with danbooru tag, using wd14-tagger with tweaks.
Additionally, every character folder contains a json file with information about training settings used for a specific model.

Installation

Paste desired model (if you want thumbnail you can also paste preview image) into \stable-diffusion-webui\models\Lora
Since LoRAs are now available directly in WebUI, you can use them as presented in the following guide.

Usage

I make models with ww prefix
some skins may have additional outfits, check lora details for name of the skin

ww_[source_name]_[character_name]_[optional_skin]
ww_ov_widowmaker
ww_al_pe_default_skin

I wanted to somehow create flexible models. I'm trying to balance my LoRAs to work at weight equal to 1, you may want to customize specific parts like hair type or length, clothes, breasts size, accessories with lesser weight if it's not working for you.

LoRAs

Overwatch

AzurLane

SocialMedia

Twitter
Reddit
Instagram