I was slightly dissatisfied with the most recent available Fran LoRA, and the old Pony one wasn't working as well as I'd like for illustrious models. So I trained one that's more aligned to my sensibilities.

This first version was trained on around 50 original comic panels from the webcomic, with some of them edited to remove text or other characters. Also like 5 or 6 fanarts that stayed almost perfectly on-model.

Activation tag: franlbx

Reinforcement tags (optional): 1girl, furry female, feline, green eyes, orange fur, striped tail

Main outfit: default-outfit, purple striped shirt, dark-grey pants

In theory, you don't need to add eye color (like "green eyes" or "green sclera") -- or any reinforcement tags at all, for that matter -- but if you're combining characters, you might want to specify each character's physical characteristics for them to show up properly.

You can also use "toony" to stick closer to the original style for some concepts or scenarios that might otherwise change the style.

Finally, I haven't tested this yet (except for Vincent), but you can, in theory, integrate the rest of the family members in an image using their e621 tags + descriptive tags like "black fur, white fur, two-tone fur, yellow eyes, white shirt, red sleeves" for Joel, "calico (pattern), blue eyes, red shirt, young" for Vincent, or "black fur, white fur, countershade face, blue onesie, toddler" for Cooper.

I plan to use this LoRA to edit more official comic panels and have more variety in the dataset to (hopefully) make a more consistent, flexible model in the future, that sticks to the original style as close as possible. And after that, I'll make an additional version integrating a bunch of fanart to allow more varied styles and representations of her. I hope you enjoy