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
Improved posing and details
Noob EPS 1.1 and Illustrious 1.0 DARE applied
FAQ
Comments (84)
Wanted to say I love all of your models and they are the best go-to, but I will say that with the latest V5 version, I have been running into many issues with it over V4, 3, and 2, but I still think its a great set of models regardless
What kind of issues? Just curious. Loving v4.0!
@Wizardegenerate mostly generations that come out with unusual fractals and geometric shapes even when i would use compatible prompts that worked perfect for all of the previous models, deleted it and redownloaded the model and it started working fine, so im not sure if it was an error with the download (rare but i guess possible) or some other error happening, its working perfect now but honestly i do also prefer V4
@daspin335 I still have v4, so maybe I wait until it improves more, until then I rather don't risk it - never touch a winning system, right?
I have a problem and need help. This is my first time using this and I'm not sure if it can work in DrawThings, because the resulting photos are very blurry. I have already used the recommendation prompts and the recommendation Settings
Use hires.fix or higher steps
Increasing CFG scale would also help
Also, I uploaded the model into PixAI as well
try it out!
Thanks,it's useful on pixai, but I still can't use it on DrawThings, probably because of the device (iPad 2022Pro)@Crody
DrawThings is very weird in several ways, sadly. I have a M1Max Macbook and gave it a try, but in the end I just prefer Forge WebUI (by alot). More settings, less drama, almost similar performance. But yeah, on iPad you need DrawThings.
I'm trying to copy some of those prompts on different models v30-50 but result is always different. The pose is the same but its like worse quality, its always looking like 20% worse, not sure why.
I forgot adding meta tag about hires.fix
@Crody are you gonna put them here? Can't seem to find them anywhere.
@jackov5162872 Same tag, 0.4 Denoise, Same CFG scale, 40*0.4 Steps and x1.5 resolution
@Crody many thanks
Thanks for the update, very based.
Oh yeah baby new version!!!!!
With the new generator update, you should totally turn on on-site generation. I can only use this model when a buddy turns on remote-share 🥺
They already are
@Crody It wasn't when I commented... But now it is! SO I'M GOING ALL OUT NOW!!!
@RedactedPaws The best setting: Text-to-Image Hires Fix
Use recommended prompt and negative prompt
CFG with 4.5, steps 20, denoise 0.4
@Crody Psh. I'm so lazy I never actually change my steps when working on an image. All I do is adjust the denoise level: .3-.4 for hires and .2-.3 for face fix/inpainting. Been working great so far. ;)
It was great while it lasted. Having all the checkpoints available was too much for them.
Btw i don't know how it works why is not available anymore?
@amogussusus Read the latest Announcement
@Crody Oh welp, seems i have to use it on yodayo again. Still cool model btw.
I've been doing some testing with the on-site generation, and for some reason Civit mistakenly thinks a lot of the LoRA models I tend to use with Nova Furry are incompatible with v4/v5 Illustrious, since they don't show up in search or can't be added in any way. Somehow it only lets me use up to v3 for them.
It seems to be an already existing issue because when I was uploading local gens it didn't let me manually add said models (that don't automatically get recognized by the system for whatever reason) for them either because the search would just show up blank. So the best I could do was to post them in the specific models' feeds. ~shrug~
I noticed the site recently added separate tags for NoobAI and Illustrious. Nova Furry v4+ changed over to that new Noob tag from the Illust one. That looks like the source of your issue.
@d_stars Well, that's going to be a problem for anyone who wants to web gen with Nova Furry now... with Civit's generator making the actually compatible models unavailable to users that way.
If you look down Setting, you can see Advanced mode on Generation Setting
Please allow it
For search, I think they will fix the issue
@Crody Well, I'll be... Took me a bit to figure out what you meant, as the 'cog' symbol in the search box is so very much missable, and I've been using this site for the better part of a year already. Good luck having a newcomer figuring this kind of shit out. Heh. =/
One of the best models here. And despite the name it is really good with non furries.
Hi. Is it just me or does Illustrious v5.0 generate worse fingers? Can you recommend something? Thanks in advance.
Including extra digits, fewer digits in Negative Prompt would do the job
Illustrious v4.0 is actually one of the most versatile models I've ever ran. It can do almost anything. It's quickly becoming my baseline for local generations. Other models do closer face closeups, but I haven't noticed any other major weaknesses.
Hi, I'm a recent user of ComfyUI, I'm trying out different checkpoints right now. The Illustrious v5.0 version has a lot of graininess, does anyone know what it's about?
Prompt: furry, anthro, detailed eyes face and fur, {Animal Type}, masterpiece, best quality, stunning quality, very aesthetic, high resolution, ultra detail, absurd, newest, landscape, 1girl, furry, anthro, solo, rabbit, dressed, body fur, brown rabbit, brown fur, fluffy, medium chest, white hair, floating hair, looking at viewer, red eyes, big eyes, worn pink basketball jersey, jersey number 03, worn pink basketball shorts, long rabbit ears, holding machete, full-length, BREAK, eyes, detailed eyes, forest in background, detailed background, detailed hair, depth of field, detailed fluffy fur, photorealistic details, volumetric lighting
Negative prompt: human, multiple tails, old, oldest, cartoon, graphic, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured, long body, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, very displeasing, (worst quality, bad quality:1.2), bad anatomy, sketch, jpeg artifacts, signature, watermark, username, simple background, conjoined, bad ai-generated
Guidance:
7
Steps:
30
Sampler:
Euler
@pavlik9425 Use lower guidance scale around 3-5
I recommend using 4.5
Also, using Euler A (Euler Ancestral) would help getting better images
I removed smZ Settings node and everything went back to normal, so I guess it has something to do with it.
@pavlik9425 did you find out why it was grainy?
nvm i found out why, i should've been using "euler_ancestral" which is euler a, i didn't know that because unlike A1111, comfy ui just named it differently ig
Great colors and contrasts, and it hits that perfect balance between promptability and creativity. World class work.
Totally bogus this got cut from the site wide generation. This was the BEST model for how low the running cost was.
EDIT: Yay it's back!
can we please have this model back
it's hard to find something good that is so helpful
I don't have the power to do it but with enough requests, they (CivitAI team) will probably consider it
@Crody i hope enough people complain so that we can get it back
i have tried some other models, but it is not the same
Now it's available again!
@Crody wuhuuuuuuuu~
Its back onlineee!! ❤
testing this tool right now.
Still love and use it (I keep V4 for now) with Forge and either Euler A or DPM 2M Karras at 30 steps and CFG 3.5-4.0. Wanted to add more observations:
- "It just works" most of the time, very nice
- While portraits 1536x1024 work great out-of-the box, details only shine when the character is close to the viewer, and for example crouching/kneeling rather than standing at full height. This leads to scenic landscape 1024x1536 renders e.g with a lying/reclining character suffer in detail and clarity. In these cases, even 1.25x upscales barely help and 1.5x has to be used to regain body details and improved eyes
- Macro/city scenes could benefit from more training (data), they are solid but still just mediocre compared to regular stuff. Examples: Cars of mixed sizes, (crushed vehicles) or (car wrecks) dont really work, (burning buildings) often leads to funny results of lava-like effects under the character's body or explosions or a 'burning vagina' etc., streets are of wrong scale, suburbia/town settings are tricky to create, scenes often lack any green/trees unless manually added with (trees), adding (utility pole) or (traffic light) makes them as tall as high buildings, etc... I am not sure what could help here, but I was surprised how good the old SD1.5 IndigoFurryMixV120 was with macro, so maybe its more of a general SDXL-based issue
- Great with anime humans (male + female!) probably thanks to the danbooru database
- I am still a bit uncertain (and up for your recommendations) in terms of step count for 1) initial render and 2) upscale render. I usually do 30 first, and then 30 again (which equals 17 steps at 0.5 denoise in Forge). Should I either increase the first, second, or both numbers for maximum details?
If the initial render can be ignored on second hires steps, you can decrease the number of initial steps to get good results
I always use 20 steps for initial steps, 40 steps 0.4 denoise strength (which means 16 steps) for hires setting
@Crody thank you, will try!
it appears that adetailer causes halos when using this model. tested on v50/40/30.
I don't understand, are versions 4 and 5 Illustrious or NoobAI?
V4 is based on Noob EPS 1.1
V5 however is based on both Noob EPS1.1 and Illustrious v1.0 which can be done by DARE merging
love the model. just wish it knew more styles
Good model.
My first results had been with the .jpg artefacts in images, but after some research I solved that problem
having the same issue. how exactly did you solve it?
Yeah how did you solve it? :(
For those asking, try changing your sampler/scheduler, that will often cause issues. DPM++ 3M SDE and Euler a are good, I use Karras or SGM Uniform.
I'm sorry if this is a stupid question, but what vae should I use?
Is it baked into the checkpoint?
It's already baked
absolutely incredible tbh
Oh damn the Illustrious 3 is not longer available is this because of the auctions results?
Yes
@Crody uhmm I notice v5 is also not on the list when you try to pick a base model on the site generator later I realize it seems to be set for noobai and not illust
@freemanx2 Try using Advanced Mode for Generation Settings via Settings
@Crody thanks for that, what a mess with all this checkpoints thing
@Crody The auction system is stupid. They should've never brought it back. Because I was literally using 3.0 because it's the only full illustrious version on this model that didn't overwrite some loras too much. And now I can't use it anymore, and forced to use the noobai one that does what I DIDN'T want it to do. The system is absolute BS
@ThetaStormer287 You could and can vote to v3 as well though by changing the model version
Just don't spread the dislikes to the system itself
Currently, I'm creating 2 versions for the latest Nova Furry XL
V6A: Extended from NF ILv5.0
V6B: Extended from NF ILv3.0
Hope you like the latter one!
@Crody Problem with that is. If I want to use v3 now, I'd need to spend an absurd amount of buzz for that. Which I, do not have. Nor do I have the money. OR EVER, intend to pay for anything on this site with actual money. Because that would be a complete waste of money to me
@ThetaStormer287 At least you can contribute and with many others feeling the same, you can get it eventually
And I know sure that v3.0 will be available again next week unless they stop auction stuff: which is unlikely to happen
The 200th model on last one got around 5000 buzz so getting at least 7500 buzz in total would be good enough
Let me use Illustrious 3.0. I was using it yesterday, why am I not able to now all of the sudden, some things get messed up if I use 5.0 instead. And I don't like how others look with my results, nor is everything I am using available on PixAi
Yeah, the pony v6 is no longer available.. da heck?
it's probably the stupid auction system. you have to bid on checkpoints with buzz to be in the generator
The bidding system is the worst and is very very unnecessary to have and I would want it to be gone honestly, plus it's an asshole thing to do for generators like this that had a lot of work and effort to not be used on the site again
P.s. it was a great when the bidding was never here but now it's just "migrate or download this to use the checkpoint if you wanna generate more" kind of bs
Why is Illustrious 4, 5, 6a, and 6b labeled as NoobAI?? They don't come out well with Illustrious LoRAs. Please fix :<
this is good i guess
Love this model, I'm not into furries, but it can do NSFW really well.
very versatile, one of my favs for anthro content
fine ig
Hi. Love this model, use it the most. But want to test other styles, like pencil or brushed for example. There is anyway to achieve this without taging an artist?
Try using sketch or line art for pencil, impasto or watercolor for brushed
Details
Files
novaFurryXL_illustriousV50.safetensors
Mirrors
novaFurryXL_illustriousV50.safetensors
novaFurryXL_illustriousV50.safetensors
novaFurryXL_illustriousV50.safetensors
novaFurryXL_illustriousV50.safetensors
novaFurryXL_illustriousV50.safetensors
novaFurryXL_illustriousV50.safetensors
NovaFurryILV5.safetensors
novaFurryXL_illustriousV50.safetensors
novaFurryXL_illustriousV50.safetensors
novaFurryXL_illustriousV50.safetensors
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.





