This Lora is trained to produce images of an OC I came up with while testing several different models and their combinations. I wanted to have a consistent element to try the different styles with, and so I came up with this character.My OCs collection: <a target="_blank" rel="ugc" href="https://civarchive.com/collections/201272">here</a>.BuT! Without further ado,<h3 id="heading-5037">This is Sivka.</h3><h3 id="heading-5038">White purple hair, purple eyes, heavy top and a surprise down under 🐶.</h3><h2 id="heading-5039">Disclaimer:</h2><h2 id="heading-5040">This is an experimental model. It was created solely as a test. Results, as with any model may vary.</h2>What I wanted to try was train a Lora on images created with AI only. Also to throw in ones, generated with an XL model and turn it into a Lora for SD. A lot of words in the prompts don't trigger/produce any result in many models, so this Lora was suppose to combine such elements and "force" them into models that usually don't generate such a result.Varying success...insert C3PO's "Machines making machines meme"

PulenKompot-Sivka [IL] | [PonyXL] | [SD 1.5]

version alpha one, bravo two electric boogalooTrained on NAI diff, adapts well to various anime models, but looses a great deal of the style and look of the dataset; produces a certain face and body type that is irrespective (different) of the dataset which I personally don't like.

Files with SHA256: 15ab40a149655a48b6003a7f6996fe99ed9eb9c25d1eecee613455153bbb9af3