Anima version is available!
Nova Furry XL
A 2d/2.5d furry checkpoint model that can have great details on any type of furs, scales and feathers, which aims to be XL-fied Peaki Furry
Rules
You cannot use the generated images for commercial use if it's not edited (or just turning it to black and white)
Advertising is always welcome
Recommended settings:
Sampler: Euler a for best performance
Steps: 20-50 steps
(Pony)
CFG Scale: 6-7
Prompt: score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_furry, BREAK, furry, anthro, detailed face eyes and fur, {Animal Type}
Negative Prompt: blurry, jpeg artifacts, username, watermark, signature, normal quality, worst quality, large head, low quality, text, error, missing fingers, extra digits, fewer digits, bad eye
(Illustrious)
CFG Scale: 3-5
Prompt: masterpiece, best quality, amazing quality, very aesthetic, high resolution, ultra-detailed, absurdres, newest, scenery, furry, anthro, {Prompt}, BREAK, depth of field, detailed fluffy fur, volumetric lighting
Negative Prompt: human, multiple tails, modern, recent, old, oldest, graphic, cartoon, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured, long body, lowres, bad anatomy, bad hands, missing fingers, extra digits, fewer digits, cropped, very displeasing, (worst quality, bad quality:1.2), bad anatomy, sketch, jpeg artifacts, signature, watermark, username, simple background, conjoined, bad ai-generated
Description
Better posing
NoobAI EPS v1.1 + Illustrious v2.0 stable DARE applied
FAQ
Comments (65)
New versions, can I ask what are the features of the new updates?
Applied Illustrious v2.0 stable and features more posing variety / characters
@Crody which do you prefer, v7A, or V7B?
going to be testing it soon
@daspin335 I can't choose since both of them are equally great
But I can say that A has 2.5D volumetric look, B has 2D flat look
@daspin335 Personal opinion of "that mouse guy" here: 7A gives prettier results by default, but 7B works a lot better if you use artist style tags, as most of those are "flat" drawings. Also better for cartoony stuff.
@Crody
Hey there, is V7A the more volumetric/2.5D model. Is V7B the more flat color/2D version of this model, like before?. Also will this still be compatible with loras that were trained on Illustrious XL V0.1?
Thanks for the clarification.
Yes they have same difference as v6A and v6B
And I think they are still compatible with LoRAs for IL v0.1
@Crody is this model inpainting compatible or do you have a inpainting model that you recommend?
@the_chucks Probably compatible
As of v7B, should we be training LoRA on Illustrious v2.0 Stable now? I feel as though the further we've gotten from v0.1 the weaker LoRA seem to be.
I don't know everything that has been merged in, but based on the version info the model is still mostly NoobAI Eps
DARE merging to my knowledge "just" tries to get the "important" (which can possibly mean new) parts of one model and applies it to the main model while attempting to not "damage" it, which in this case seems to be NoobAI
Also as a side note, it'd be hard to get any good e621 knowledge out of Illustrious v2.0.
im getting black screen when generating with 7A/7B, anyone know what might be the problem?
It's either VRAM outsourcing, NSFW filter or corrupted file
Try using aria2 for downloading
@Crody VRAM outsourcing? What exactly do you mean by that. I'm having the same issue of the images just coming out as black or corrupted somehow I just did a fresh install of A111 and this model.
@Disaccountdumb Some speculate that it caused by the system itself: A1111 webui
Since I can create images without having any issue using diffusers+sd_embed (which is much classic way to do t2i), the black image can be happened by the differences between these two systems: VRAM usage
Assuming the system is local, did you download it via normal downloading?
Try using aria2
@Crody I downloaded it normally yes but my System has 16GB of VRAM which is why I'm confused if its using too much. I had it working before on this exact same system but now it doesn't after I had to reinstall everything.
I've had black screen issues with A1111/Forge, usually when I put too much in the prompt and use lots of loras (I'm a 10GB peasant) but flushing the VRAM usually fixes it for me. Never had the same issue with ComfyUI though so I mostly use that instead.
What is the token limit of Nova Furry XL?
Same as other SDXL models
75 tokens for classic systems (unmodified Diffusers for example)
Almost infinite for modern systems but I'd recommend around 300 tokens for the best performance
@Crody Is this for Pony? I can't remember if I ever made Illustrious/Noob collapse under heavy prompts
@Stinkek Yes as well
Okay, I need to know of the newer versions (7a, 7b, 6a, or 6b) which is closest to v5?
Hmm need versions
<3
Do you know what this model needs and in general the vast majority of furry models??? Cow women, I don't understand why no model I've used has good anthro cows, whenever I want to make a furry cow they come out with too many human features even though I indicate it in the Negative Prompt, and very rarely do they make the udders, I don't want a human with cow features (ears, tail, horns, etc.) I want a damn anthro cow.
I don't share your tastes, but I share your frustration. AI models can't do rabbits or rodents without a prompt ten times larger than it needs to be, and even then they arbitrarily screw it up half the time. I'm convinced negative prompts don't actually do anything in SDXL either.
Someone get this rancher some AI heifers. S/he deserves 'em.
I've never had an issue genning cows
https://files.catbox.moe/1toses.png
The Illustrious v.30 needs bidding what do I do?
BID ON IT??!!
what is the difference between 7B and 7A?
@Crody thank you
Anyone know the best way to prompt that gets rid of the like "belly patch"? Like I'm trying to gen a dog character that's all orange, but it keeps giving her a lighter tan strip from her belly to her muzzle. How do I make the fur all one color?
can this model work with controlnet?
She has such amazing eyes for a fox.
@Crody What's base model should we use to train loras for Nova Furry XL?
Illustrious-XL v0.1?
Nova Furry XL has its own unique encoder which is mixture of NoobAI EPS v1.1 and Illustrious v2.0 stable
Using Nova Furry XL for training is the best option and the LoRA should be compatible with Illustrious v0.1, v1.0 and latter and NoobAI EPS
This may seem stupid, but how do you make the bellies bigger than the characters body like 2x or maybe more bigger in this model?
Thank you if you do help ^^
Try adding tags like "hyper pregnant" or "belly inflation" and increasing their strength if it's not enough.
Truly the model of all time. (This is a compliment)
Goated
Has trouble differentiating arms from each other; it will decide on a case by case basis whether it'll affect the right arm, the left arm, or both with tags like 'prosthetic'.
The only problem I have is the sheath, I don't know why, but a lot of time, it changes the sheath for a weapon sheath that is annoying.
This issues has been reported already and sadly still didn't seem to be "fixed".
Very annoying, as sheaths should be more than plenty enough in the database to train on e621.
Which website was used to train this model?
has anyone converted this model to ONNX? If so please let me know, its so hard. :(
Is this compatible with with stable diffusion webui forge or is there another locally run application one can use for this checkpoint model?
Any kind of system can run this model if the system is available for sdxl models
For this case, forge is available for sdxl so yeah it's compatible
I also use Forge, no problems! Euler A - Automatic, CFG 3.5-5.5, around 30 steps, rest up to you.
Anyone know the most reliable way to gen a female character without hair? I've tried neg prompting hair, pos prompting no hair, etc. Is there a specific term I'm not thinking of?
The ‘bald’ or ‘bald head’ token works for me in most cases, but it is not consistent.
using 'bald female' usually works, try in combination with 'hairless'
Amazing model! May I ask what image size you recommend for generation and what size for upscaling?
I suggest using 768x1344 or 768x1152 for portrait
And then I upscale it with x1.5
@Crody thank you so much!
Personally, I do 1024x1536 (or reverse), or 1280x1280. Works really well as well, little more detail
Agreed, go straight with 1024x1536 and then upscale by whatever you like (my system is rather slow, so I do either 1.25x or 1.5x). Going to 1024+ for the stock width (x-axis) size helps with details a lot. In general I noticed that it is very robust and almost glitch free even at that rather high stock resolution. only with 1536x1024 (landscape) it tends to produce some more errors/duplicates. Though that can be used to create duo characters / couples.
Mine are usually 912x1144 then upscaled 2x
Fantastic model 10/10
Amazing model!
In some cases, this model is smarter than Nova Animal V4. I do not know why, but OK...
For realism models I think they kind of have to destroy some of the creative ability in order to make it look more realistic, either that or the models being merged just aren't as well trained on specific details as this. Illustrious base itself is an anime model too :(
@furproxy I thought the same thing.
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.