IL v2.3 : New version, more flexibility. Better than IL v1
This version don't have a baked in VAE. I recommend to use SDXL VAE
Please consider supporting me so I can continue to make more models --> https://www.patreon.com/BB95FurryMix
Don't forget to join the Furry Diffusion discord server --> https://discord.gg/furrydiffusion
Since v3, this model uses e621 tags.
This model is a mix of various furry models.
It's doing well on generating photorealistic male and female anthro, SFW and NSFW.
I HIGHLY recommend to use Hires Fix to have better results
Below is an example prompt for the v7/v6/v5/v4/v3/v2.
Positive:
anthro (white wolf), male, adult, muscular, veiny muscles, shorts, tail, (realistic fur, detailed fur texture:1.2), detailed background, outside background, photorealistic, hyperrealistic, ultradetailed,
Negative:
I recommend to use deformityv6 and boring_e621_fluffyrock_v4, you can add bad-hands-v5 if you want
Settings 1.5 :
Steps: 30-150
Sampler: DPM++ 3M Karras, DDIM, UniPC or Euler A
CFG scale: 7-14
Size: from 512x512 to 750x750 (only v4/v5/v6/v7/v8/v9), up to 900x900 for v10/v11/v12
Denoising strength: 0.6
Clip skip: 1
Hires upscale: 2
Hires steps: 30-150
Hires upscaler: Latent (nearest)
Settings XL :
Steps: Below 30
Sampler: Euler A
CFG scale: 3-6
Size: 1024x1024
Denoising strength: 0.2
Clip skip: 1
Hires upscale: 2
Hires steps: Below 30
Hires upscaler: Remacri
Description
This version improves fur and will be great to generates better bodies
FAQ
Comments (61)
With each update there is noticeable changes with how shadows and other lighting related things have improved. from v13 to v14 the lighting on the faces seems to be the main improvement besides higher quality sample training. Its great to see the bb95 still being updated.
Umm... I have a silly cuestion. Can I use this checkpoint for making anime pics? or hybrid style like anime + realistic?
yeah it's possible ^^
You can, but make sure you say they are a "Human" near the start of the prompt otherwise they may have mouse ears or cat noses or tails and stuff like that.
@tombot Of course, if neko-girls and the like are what they're going for, then hey, no problem there! :) (Disclaimer: I've never tried to make such things in this or any other model here, so I could be wrong. ^^; YMMV and all that.)
@tombot Nice! But I mean furries in anime style, like Yohei from Grimoire of Zero, or the Wolf Trio from Dragon Ball Super.
Thank you everybody for answer me!
@teren0894 it should work yes, i did make a Yohei lora that i tested on this model and it was working well ^^
I love this models,thank for share
Don't know how you do this but you really keep improving this model! So far every version I've tried was better than the last one. Also I think most people underestimate its img2img powers. In my opinion the absolute best Furry model for img2img by FAR and I also absolutely love the soft fur you can achieve and its distinct realistic style. Thank you so much for working on this. A true gem.
Absolutely amazing model. I hope you make an XL version someday 😁❤️
Maybe one day ^^
My only feedback would be that human eyes in this regularly appear as basically just empty white blobs, with no pupils or anything.
Will V15 be photorealistic and perhaps good for female hyper muscles?
This is so cool! Thank you for creating such a cool model, so much room to create different styles
When using Hires upscaler: Latent (nearest), some digital noise sometimes appears.Help?Or can I use any other?
@VladossX Latent-based upscalers are known to act weird especially when the denoising strength is set out of the best range. You can use this video as a short guide https://www.youtube.com/watch?v=sre3bvNg2W0 , the diagram shown at around 07:09 features the Denoising Strength ranges for various upscalers. Maybe also try the 4x-UltraMix_Smooth upscaler as well? It's mentioned in the description of the YiffyMix model. Cheers!
@oaf40 I just started using "R-ESRGAN 4x+" because it sounds cool.
What do you use?
@VladossX I don't use upscalers on my own computer because I don't have enough VRAM to use any. On occasions when I generate pictures on rented GPUs (Vast AI) I use either R-ESRGAN 4x+ or R-ESRGAN 4x+ Anime6B. My personal preference is the latter one because it keeps the outlines sharp.
Works great
Absolutely love the way muscles and scales are done on this!
if you plan to make sex scenes, DO NOT waste you time on this one. It's just godawful with sex scenes. Can't do $hit when it comes to anything more complicated than your average "doggy style". Also, can't handle "male pov" properly
@tpcdaz I've yet to see any sex scenes pics gen'd with this chckp at all lol. How do I know? There's a pretty large dis group for AI furry arts. None of them even use bb95. Even more, I've tried to just copy-paste the exact prompt and gen info of pics (there's a feature you can scan an image with, giving you whole data on the pic should the poster leave the PNG info there), and bb95 end up effing it up even after this.
again, its okay when your prompt is just like ten words long. The problem starts when you want some unusual sex poses, or want to describe your chars more presicely
It's possible, however as with all models it can take some editing luck or and like regional prompting, some poses like the top's POVs are easy enough without much work ¯\_(ツ)_/¯ otherwise well, what did you expect? the Ai in general can't get the number of fingers right, and im not even talking about holding objects or interactions between characters.
@yiffysan So what is suggested
@Weoxur I suggest using other models for sex scenes
@yiffysan Which ones though, I barely know any good furry ones, Indigo furry mix?
@Weoxur lots of good sex scenes I've seen are made with autismmix. You don't even have to use "furry exclusive" chckpnt. A good one can do it hentai/furry style, just have to make a proper prompt
@NCDarkness_ "there's plenty of LORA for specific poses" .. and the 80% of them are made for XL chckpnts lol. good luck with genning a human male pov with a girl wearing pantyhose grinding the pp so the pp wouldn't look like it was cut and just put underneath the girl
@yiffysan lmao the way you explained PNG info tab of A1111 like it was some mystic thing we might not be aware of immediately tells me you're just maybe not good at prompting lol.
@awoo effed up logic lol. but whatever really. I like the unaware people who literally have no empathy. And since I do have some, it's not too hard for me to speak up details so everyone understands what am I taking about since they can be new here. While the unaware people think everybody is experienced af
@awoo it can't even properly make a pov shot with a girl lying on front, plus just human hands pulling down girl's panties so the anus is only visible lmao. What are we even talking about?
@tpcdaz and the winner of the "most toxic m0f0 of the year" prize is.... ta-da! tpcdaz! congratulations! I 146% bet you haven't even tried to gen anything other than solo or some starter pack "doggystyle" pics with that checkpoint. If even any at all
@tpcdaz nevermind. Checked the pics you've posted in your profile.. Just as I thought. Not only there is none made with said checkpoint but not even any furry one, just some basic solo pics made with easy af realistic checkpoint. "That said it all"(c) And you still even dare to call me out on handling AI lmao. what a clown
@tpcdaz It's pretty sad. If you don't like the checkpoint literally just use any of the billion other ones out there. I have to wonder what has to happen to a person to make them act like this.
@tpcdaz blah-blah-blah. Why don't you make exactly a scene with what I've desribed and show "how it's done" lmao, without using lora, inpaint, or sketch. I double dare you. Btw the only prick here is you, agressive m0f0
@tpcdaz blah-blah-blah. All talk and no business. Why am I not surprised
@yiffysan Look, here's what I don't get. You clearly don't like this model, you've claimed it's "godawful" and "can't do shit" and seem to believe it's fundamentally broken somehow.
But on the same note you want people to tell you how to prompt the things you want or just "prove" that the model can do them... why bother? Why do you keep coming back? For over half a year. Just move onto another model that works better for you. Even if this model was limited in the ways you claim, why the need to attack people over it? What does this achieve for you?
The only logical thing I can assume is you actually *don't* think the model is godawful, there's stuff in the gallery you really like and wish you could reproduce but somehow can't, and instead of asking nicely for help you just insult everyone. Talk about the fox and the grapes.
I think it's clear at this point that arguing with people is far more important to you than generating AI art or even explaining your thoughts cohesively, so I expect you'll just reply "I have more empathy than u noob and fk ur dum model" and convince yourself you owned everyone again.
@awoo as I stated in the first place, it's godawful with sex scenes (if you want anything more complicated than avearge and boring af "doggy style", etc). Not with solo pics. With solo pics, it does great.
Me attacking people? wow. That errogant m0f0 has started it, trying to look wiser than he is. So then I dared him to gen a said sex scene not to get a free prompt from him (cmon, the prompt is obvious), but to curbstomp his ego. Because I know for sure he would try to gen said scene (such an errogant overconfident m0f0 couldn't not to), and then realize he can't without using lora, or inpaint
@yiffysan Buddy the only person here with an ego is you. Nobody needs to prove jack shit to you.
You also made the exact sort of reply I said you would. Go get therapy.
@awoo @tpcdaz
Yiffysan isn't wrong guys. The checkpoint on its own with a single prompt can't do it.
Denying that reality doesn't make the problem go away and doesn't help bb95 with improving anything.
You shouldn't have to regional prompt, inpaint, openpose etc to get two characters doing what you want them to do.
@tpcdaz and what were those "sex scenes without loras and inainting"? Some boring as fuск plain doggy style? (which I can get myself without a problem). Wow what a breakthrough. I see you can't read. Well, for some special m0f0 like you, I'm gonna state it again, slowly, so you can follow. It. does. those. common. poses. without. issues. Now put some imagination (if your special mind is capable of, ofcourse) and try to make some really unusual sex scenes without using loras and inpainting, and then stfu when you fail at it
@narwhaleforsale480 I bet those two just never tried to do anything but some usual boring doggy style/cowgirl. That's why they think it does any sex scene without a problem. Also, they can't read lmao, since I've stated BB95 does those common poses okay, and the issues start when you want something special
@yiffysan exactly 100%
@narwhaleforsale480 LMAO you made a sockpuppet!
@awoo yep. I've created that profile in the may of 2024, and was saving it until now.. (facepalm)
boy it's really outstanding how some are so delusional they come up with such BS just for more doses of copium.
Ironically, you DO know bb95 can't do $hit when it comes to sex scenes any more complicated than default doggy/missionary/cowgirl. But somewhat you feel like keep making that gaslighting manipulation "bb95 always does everything great, git gud". Talk about needing a psychiatrist
Awesome checkpoint model. I use this, Yiffymax and Indigo. Though your model doesn't seem to give tails when clothing is worn. Even with prompts like; tail clothing and tail entrance etc
Those prompts don't really make a lot of sense to the AI. Just try weighting "tail" and other e621 tail related tags higher.
@awoo You're right on that one. ^^
I don't know if it's only me, but through the civitai generator, the specific model has been showing problems with generating and often doesn't generate images. It has been there for a few days, saying it has those problems, and it does not work. Maybe if it was updated (without necessarily changing anything) that could fix it?
Old versions were very versatile and supported countless disparate styles. All of them were seemingly trained out with each new version, making the output and model awfully homogeneous, way less useful and a lot more boring.
My recommendation: Stick to something like V7 for more artistic possibilities.
In its defence, this model was always highly focused on making photorealistic style furry art, like in the creator's example images, and it has only gotten better at that with every update. That obviously comes at a cost. I'd personally suggest if you want a more "general purpose" 1.5 furry model with a greater understanding of different techniques and artist styles, try Indigo Furry Mix.
I haven't done a lot of 1.5 stuff lately as I'm focused on Pony, but when I was I'd generally go to BB95 when I wanted consistency over different seeds and photorealism, and Indigo for just about everything else, especially heavy LORA use. If you do try Indigo, my personal recommendation would be the vpred version, but hybrid is also great with artist styles.
Try a neutral prompt
Most of the later models for the later versions of A1111 take more elbow grease than they did in the past.
Have you thought about doing an SDXL variant of BB95 Furry Mix? (Potentially based on Bananastrike XL Lightning) I think it could be a success :)
Yeah, I'm working on it ^^
@Bumblebee95 There's a better option: NoobAI-XL.
I am strugling making futas with this checkpoint actually can the autor please adress it in the next version?
so its just me or the model was deleted from civitai generator?
Was wondering what was up when I went to generate just now, but yeah I don't see it at all.
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.



















