2D Focused Furry Illustrious Model... wowzers.
Illustrious model focused on 2D, traditional, digital- all that stuff, like 90% of other Illustrious models in existence. Not a realism model, can't (-ish) do it, and I don't intend on this model going down that path, that's what this is for.
Model information and details continue below, read before use.
Links are provided below for those who can't run this resource locally.
Version Notes Below:
Version 2.5 release.
Re-centering of model style and conceptual knowledge, i.e. going back to 2D.
Furry concepts should be improved, they seem much improved to me, but then again I have no idea what people use this model for so who knows. I nuked my PC doing this, obliterated my main UI, and had to reinstall my drivers after making this thing. I'm a little frustrated.
Here's a prelude to that:
But here's the results. A version increase.👏👏👏
Wowzers... changes follow below.
Changes
Improvements
Improved knowledge
Better light gen
"AnGleS"
'Better Betterments for Bettering Better Images ™'
Also, tuned for DDIM, Euler A, DPMPP 2M/3M, I would recommend DDIM with Uniform for more muted image colors.
I actually had more on the list for the version, like increasing gen resolution to reliably do 1536x1536 and re-introducing the more esoteric styles/framing that I typically gen with (but never post); I just didn't want to spend another two weeks working on it just for it to not come out. This model is technically the first-half of a two-part mix that I haven't mixed (yet). I do have a model that performs as I wish, that actually does what I stated prior. It's just not a massive gob-smacking increase like people would expect.
<About The Model>
It's an illustrious furry model, designed to output furry content. That's not all it can do but it's where I placed my focus while merging, along with character accuracy. It understands the majority of artists' styles to varying degrees and most concepts. Compatible with all Illustrious Loras and many NoobAI Loras, though use of those aren't necessary unless you're prompting for very specific niche concepts, characters, or certain styles. I don't show much of what I make (no one wants to see 50,000+ images), but I have posted some images to reflect some, but not all, of the model's capabilities.
Wish to use this model but don't have the hardware to run locally?
You can run this model for free on the sites linked below (and here):
(Remember to read and follow the ToS of each site.)
SeaArt.AI
https://www.seaart.ai/models/detail/5548eaa0088cef536fb6cacb88cc038a
Tensor.Art
https://tensor.art/models/962141417671531835/Illustrious-Lima-V0.2.5
PixAI.Art
https://pixai.art/en/model/1829757967386098908/1976038955748236033
Yodayo
If there's a model hosting website that I don't know about that you use and wish to generate on, let me know and I'll look into providing the resource on the site.
If you wish to use this model on this site (Civit), bid. I'm not throwing 15$ at this model every week just for people to not use it.
Alright, cool, V2.5 is out. I had about this much fun making: . Amazing right? There's nothing there! Anyways, next version, because believe it or not this isn't what I originally wanted as a V2.5 release, it's just how it worked out. I have things lined up for this model such as: making the model more responsive to non-artist prompted styles, and, improving base 1536x1536 generation outputs. I actually already have a model ready that I mixed for specifically that purpose, but legitimately, who's gonna use it? Despite the limited goon-slop I post, I actually don't use my models for just straight wankery. I'd post what I actually gen, but Civit ToS doesn't permit gore and acts of self harming/suicide or implied acts of such.
<Usage Tips>
CFG: 2 - 9
Steps: 20 - 50
Sampler: DDIM, DPM, Euler, I recommend DPM++ 2M/3M but Euler A is also perfectly valid.
Scheduler: Beta, Karras, Simple, Normal, Uniform is recommended (Sampler dependent).
Resolution Size: 796 - 1,700, the model excels within this range. When exceeding the upper range, prompt intelligently and use the resolution space wisely to avoid image subject deformity at very high resolutions (1,900+).
My Settings
My recommended settings (settings I generally use):
CFG: 2.5 - 5 (Higher chance for monochrome/sketch style at lower CFG's, I would suggest using the negative prompt to counteract this or prompting for style like normal. When it comes to style prompt, [i.e. actual artists] I would recommend CFG's around 2.5 to 3.5, but if you like overbaked images 7.5 is good too.)
Steps: 35 (Sometimes I go higher, sometimes lower. I typically go for 50 on other models because that's just how I gen. That's not needed with this model, 35 is a good baseline)
Sampler: DPM++ 2M
Scheduler: Karras or Automatic, Beta if Euler A
Resolution Size: Angle and pose dependent
Use however you please, feedback is appreciated, crediting is unnecessary.
Feel free to post whatever images created in the gallery below.
I'm not responsible for generations produced with this model, generate responsibly.