Why this LoRA exists:

So, I couldn't find a LoRA for the WTFeather style, so I decided to make my own. Ta-da, here it is!

Strengths: LoRA has absorbed the original author's style well. The feet are pretty cool, lol.

Weaknesses: WTFeather's fetishes are pretty... specific, and this has left its mark on LoRA's work. Almost all the pictures in the dataset depicted girls drooling and crying, with their mouths wide open in laughter, and LoRA tries to follow this pattern.

Instructions:

Main tag:

wtfeather style

This encourages the model to mimic the WTFeather style of drawing. For better facial expressions, I recommend adding a couple of tags like

calm, shy smile

or whatever comes to mind.

Additional tags:

sybian, sitting on sybian

Do I need to explain what this tag does? To my surprise, the sybian works almost the best in this LoRA.

feet_wall

Well, it's a WTFeather glory hole with feet sticking out of it. The tag itself works pretty well until you try to specify who the feet belong to, lol. Seriously, it works more or less adequately with an abstract character, but if you try to specify that the feet sticking out of the wall belong to Jinx from Arcane, everything breaks down.

meat wall

I didn't know what this concept was called, so I made up the name myself. When you add this tag, your character will get stuck in a wall of pink flesh. Best used in combination with

tentacles

tickling

What can I say... meh. Disappointing. LoRA barely replicates the original author's signature fetish. If you still want to try, here's how it should work (in theory):

1. Use the "object tickling" tag to select the tickling method (hands/brushes/feathers/etc.). Example:

hand tickling, hairbrush tickling

2. Use the "bodypart tickle" tag to select the tickling location. Example:

armpits tickle, feet tickle, navel tickle

3. Try it, experiment, discover something new, because I have no idea how this thing works or if it works at all.

bdsm, bound, femdom, yuri, arms up

Just tags that work well.

Settings:

DPM 2 Karras, CFG 5-7, Steps 30, resolution 1024x1024 or 1216x832 (in any orientation) - with these settings I got, in my humble opinion, the best results. I also liked Euler A Automatic.

Future plans:

This LoRA hasn't even been released yet, and it's already asking for version 2.0. Well, I'm going to work on it, but first the following resources need to be accumulated: Buzz for training (easy) and an improved dataset for better training (hard). In short, the more images you generate using this LoRA, the closer the birth of its second, improved version, without randomly open mouths and sudden tears.