ILL v2:

Trained in OneTrainer at higher steps and epochs on the same data as v1. This should fix some issues like the spots not being filled in, and the hair should be more accurate (less detailed, but closer to her locks).

For some reason, the AI likes to add brown fur to feet and hands, so add

negative prompts: black fur on hands, black fur on feet

Z-Image v1:

Trained in Ostris AI-Toolkit. Samantha, character by Jay Naylor. Originally tried to train on the datasets from the comics Samantha and Dogging, but it wasn't working out, so this is trained on AI-generated images from my V2 Samantha Illustrious LoRA instead.

I don't know how to really prompt Z-Image, and without ControlNet capabilities or a finetune, it is a gamble on good results unless it's basic. Will revisit when the base model is released and a good furry finetune is created. This is just a test of Z-Image as well as familiarizing myself with the Ostris AI-Toolkit program.

ILL v1:

Trained with the onsite Civitai LORA creator. Training material just pulled haphazardly from the Jay Naylor Comics: Samantha and Dogging. Works best with NovaFurry Illustrious.

core tags: samanthajn, bobcat, short tail, furry female, collar bell, grey hair, blue eyes, ponytail

add these tags if the spots on the body aren't consistent: spotted fur, spotted body

other tags: dog bone earrings, black jacket, fingerless gloves, black boots, tank top, black thong

negative prompts: text bubbles, speech bubbles, text