Florans - Starbound [PDXL]

6/14/2024 PDXL Update:

I updated this model for use with PDXL. Please see the "About this version" section to the right for details. It's pretty much the same model, just trained on better tags, with new images, and at a better resolution, and with a better optimizer. Dataset was expanded by 58 images, so it's now 243 images! Technically, PonyDiffusion is already trained on an okay amount of Floran art, but was mostly incapable of generating them successfully, so this lora will help a lot with the more tiny details like their eyes and mouth. Hope you enjoy!

-To successfully generate their eyes, consider using "black eyes" and "empty eyes" as tags in your pompt, without empty eyes, they might have little white or black dots in their eyes, imitating pupils.

-For their hair: Make sure to use "plant hair" and "pseudo hair", specifically the pseudo hair tag helps a lot with getting their plant-like hairdos right.

-For their mouths: Using the tag "mandibles" can help pretty greatly if their with little fangs aren't showing up. I would avoid using the tags "teeth" and "fangs", though, as those usually just add white teeth rather than their weird little lip-fang-things.

-Florans come in all shapes and colors, so consider using wacky color combos on the skin and hair! There is a definite bias towards green skin, so make sure to prompt against that if you want anything else. I recommend using a wildcards file for randomize colors for fun.

-This model has a few traces of of Nuru from the game, however my attempts to actually generate her have failed, so I don't think this model can generate her

Older Description:

It's the Florans, from Starbound!

This is a lora trained on a whopping 185 hand picked images of the Floran alien race from Starbound! It was trained on exclusively digital art. It wasn't trained on any specific floran, but just the race in general. I'm quite satisfied with the results of this lora, mostly because of the variety you can get from the characters. It was trained on e621-style tags, so use those when trying to prompt - I've included the tags in a .txt file in the "Training Images" download below.

It works pretty well with most .ckpts, however, since it was trained on e621-style tags, it might be ideal to use a model that is trained on a furry art. I would recommend using LucidMix, or Lawlas's Mix. They gave me great results and I used them for most of the example images.

NOTES:

-Naturally, florans are a more tribal-kind of people, so "tribal, tribal clothes" will yield relevant results.

-The eyes can be a little inconsistent depending on if you want anime-style eyes, or the pure black floran eyes. If you want all black eyes try "black eyes, solid eyes" in your prompt

-It was trained on a fair amount of male florans, too, so it can do those just as well as female florans.

Example prompt: (See example images, all images should still have their metadata. Just drag the images onto the PNG info tab and import it.)

It smells like fruit in here...