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
Enhanced posing and higher contrast
NoobAI EPS v1.1 + Illustrious v2.0-stable DARE applied
FAQ
Comments (30)
Can I access older versions?, versions of novafurryXL after v110 do less stufff than earlier versions
You can use them on other sites such as PixAI and SeaArtAI
Links in the description
@Crody thanks for the quick rpely but if is still the newer version, it doesnt help :C
@jmfmagnum32959 Click on Show More, Scroll down, and you'll see other versions like V10 or Pony versions
@Crody oh....thanks wish i could download those versions just in case, though I have no idea how to use them haha
@jmfmagnum32959 You can change the version to download via this site by clicking on the version on the tab under the tags
Under the three dots and the hammer, over the showcase images and the word Download would be the location of the tab
@Crody wow thanks again, sorry for bothering
Less stuff? For example?
@Raylom more exotic scenarios and outfits like skin-tight space suits or a diving helmet
@Raylom forgot to add in ealier versions you could even turn the image 3d, was cool
Would it be possible to put a quick description of the changes from version to version? Like what improvments/fixes/alterations have been made with each new one that comes out?
This would be very useful
V13.5 and V14 were a big letdown, so I had low expectations for V15. Well, as it turns out, V15 is a great model! The anatomy is much more consistent, flexibility is back and it looks much better than V12 visually speaking! I was about to move to Yiff in Hell because I didn't think V15 would be any good, but it is! V15 will be my main model now!
I'ma be honest I can't tell what improved. We need an update log or something
@123inactace Turns out i'm fucking stupid
V15 is awful as hell, man. Prompt cohesion? Barely heard of it. V15 attempts to add human males to every girl i'm trying to generate
@PolskaPshe Could you provide some prompts / settings for those generations if possible?
I think prompts like solo isn't added or trying crowd into the negative would help reducing those issues
well damn, im holding onto 1.35 as much as possible, but the site im using soon will update it
@Crody problem solved. Looks like i'm using too many prompts like "anthro" and "kemono" (3 of each lol). Still, style cohesion is nonexistent
@PolskaPshe
Style cohesion can be achieved by stating some prompts / negative prompts revolving the style
For example, the thumbnail image (fire fox girl, Vulpyra image) has several prompts that revolves around the style like 'anime coloring, high contrast, volumetric lighting' and negative prompts like '3d, rendered'
I recommend taking look at some of the example images' prompts / settings in order to use it with the best condition
Is there a suggested way to eliminate human males specifically? I've always been able to brute force it out eventually, but has always been the one thing I've struggled with, especially on gens when the male is only partially in frame.
You can add human, male to the negative, and solo, furry, anthro to the prompt for the reducing
Also for the CFG, I recommend using 4 to 5 instead of higher numbers
Higher than those would create unwanted parts to the image
@Crody Yea, I run most of my gens at CFG of 3. To some degree, I know there's always gonna be some brute forcing when dealing with multiple characters. Realistically, my gripe is probably just an overarching issue with some angles/positioning/etc. just inherently having stronger biases.
@Disarmony007 I recommend checking some of example images' prompts and various settings
Mine would be:
CFG: 4.5, Steps:20, hires scale: 1.5, denoise strength for hires: 0.4, hires steps: 16
V15 overcooked as hell. All the unique artstyles went out the window and all it can gen now are just "1girl masterpiece best quality" AI style images. Artist tags that were working fine in V12 now does nothing. Change your training recipe bro. Throwing the same cake back in the oven for the 15th time doesn't cook it more just burns it.
Must Ask question: have plan add for model Bakugan ?(Ask but need creating this dragoness without using lora)
I'm not sure, v15 kind of produces overcooked results, as another person said.
I feel that v11 was the best for me, so I will downgrade and wait it out.
By the way, I have a question. What is the recommended VAE to use with this model? Which VAE do you people use? Do you even need one at all?
👌i agree somewhat so i decided to download both versions and run them separately when I feel like one is lacking over the other. Most of the time the v13 or v11 (earlier) versions work better than the later versions with some LORAs, so you will get an entirely different result depending on the LORA used.
I compared V14 and V15 and the big issue I have with V15 is that it tries to add things that isn't in the prompt. Let's say I just want a picture of a furry cat with a fully white fur. With V14, most of the time it gets me what I want, and not so frequently does it gives me a picture of a mostly white furred cat with a little bit of added colors like orange ears, grey hands, etc. But in V15 it does that way too often and I could barely get the picture that I want. Adding "completely white fur" or increasing the weight either doesn't do anything or just fries the image. And while this adding unnecessary details issue isn't as severe in V14, it starts to get annoying when I want to generate a blank t-shirt or a blank jacket as it tries to add decorations like a logo, picture, emblem or writing on the clothing item. And adding "blank t-shirt" or "blank jacket" barely helps. However this issue very rarely happens in V12 making it the best NovaFurryXL checkpoint so far when it comes to prompt adherence, but I barely use it because I love how balanced the details and lighting/shading looks on V14. Hopefully you'll release a version that fixes that!
V14 is just perfect talking about lighting, shadows and etc, but v15 feels more heavy, (overcooked) sometimes more doesn't mean better guys