Update 19.06.2026. Lustify V9 (Zenith) IS OUT!
This is the absolute final version of Lustify built on the SDXL architecture. You can find the full changelog and a breakdown of the new features there. <----------
Recommended Workflow.
Alternative Access & Support: Outside of CivitAI's Early Access, you can get the model via my Boosty.
Just like last time, after the early access ends, there will be a 1-week period where the model is available for "on-site generation only." After that, it will become available for everyone to download.
Alternatively, you can make a direct Bitcoin (BTC) donation, just make sure to DM me after that: 1CoMZvrXtRXbz7BCzdvxifSjHqhaf1yUZs
LUSTIFY
Lustify is a photoreal checkpoint optimized for creating high-quality imagery of women in explicit/sexual scenarios. The model has been trained on a hybrid dataset, meaning it natively understands both Danbooru tags and natural language prompting.
Note: While it has a strong NSFW bias, if you guide it properly, it is fully capable of generating stunning SFW photography, objects, animals, and fantasy environments.
Crucial Points Before You Download:
NOT Pony or Illustrious-Based: This is a standard SDXL checkpoint. Do not use Pony/Illustrious LoRAs with it; they will not work correctly.
Distant Shots & Anatomy Limitations: Like 99% of SDXL models, distant shots can sometimes suffer from distorted faces or hands. This is an inherent limitation of the base technology (4 channel VAE), not the finetune. It is incredibly easy to fix using Inpainting or ADetailer. Please utilize these tools rather than reporting this as a bug.
Experimental Nature: My training workflow involves a lot of unconventional, high-risk experimentation. Because of this, some prompt combinations that worked perfectly on other checkpoints might behave differently here. If you encounter unexpected behavior, feel free to give feedback so I can look into it.
Recomended parameters:
Sampler: DPM++ 2M SDE/DPM++ 3M SDE,
Scheduler: Exponential/Karras
Steps: 30
Cfg: 2.5-4.5.
Highres.fix: upscale by 1.4-1.5, denoising ~0.4
For the distant shots of people you probably wanna use highres.fix and/or adetailer.
Also, use reasonable positive/negative prompts, shizoprompting does more harm than good (with any SDXL model).
Some tags that have a big visual impact with this model:
Camera type tags: "shot on Polaroid SX-70", "shot on Kodak Funsaver", "shot on GoPro Hero", "shot on Canon EOS 5D", "shot on Leica T".
Photography styles: "analog photo", "glamour photography", "street fashion photography", "candid photo", "amateur photo".
Lighting types: "cinematic lighting", "neon lighting", "soft lighting", "dramatic lighting", "low key lighting", "bright flash photography", "warm golden hour lighting", "radiant god rays".
Film types: "Ilford HP5 Plus", "Lomochrome color film", "Fujicolor Pro".
Photographers: Alessio Albi, Martin Schoeller, Miles Aldridge, Oleg Oprisco, Tim Walker.
Others: "film grain", "bokeh", "dreamy haze", "technicolor", "underexposed", "low quality", "lowres"
Alternatively, you can give a read to this wonderful prompting guide by PromptGeek. It has A LOT of useful tips I didn't know about even after years of using SD.
Description
FAQ
Comments (157)
prompt a girl from side please with your latest model and post it.
You just improved my whole Sunday🤩, gotta test this asap
You are the best! Thank you for another round :)
I'm downloading it and can't wait to try it. You're doing great work. Respect for effort.
Any tips to avoid having a smooth face? That’s so frustrating, haha! (body is perfect)
Super that you are still developing the model, it's my favorite. I didn't fit a couple of details in the previous version so I took it in and reworked it a bit in my style. I was going to post it soon but now I'm going to hold off a bit and play with your new great creation
I did a quick comparison. Your new model is excellent! https://civitai.com/posts/13956298
Babe wake up new lustify dropped
Excellent! Are there any plans for DMD2 version of OLT?
You can load the DMD2 Lora alongside it.
@MachineMinded yeah pretty much. It's not that hard tbh.
Nah, no point. It's literally normal checkpoint merged with DMD lora
Here is the DMD2 download link: https://huggingface.co/tianweiy/DMD2/blob/main/dmd2_sdxl_4step_lora_fp16.safetensors
@coyotte what settings for your merge if you dont mind sharing?
@darkmatter151 You simply merge it 1:1 with the DMD2 lora. Or add this
<lora:dmd2_sdxl_4step_lora_fp16:1.0>
in A1111 or Forge.
I was really looking forward to the new model, but the first tests showed (photorealistic images) that the detail of the new model is lower, and it takes several generations to generate a decent image compared to one in the previous model. It seems that the generated ai photos used for training had low detail and had a bad impact on the final quality :(
yes, this is my theory too.
amazing as usual
Downloading now, thanks for your efforts! Endgame was the best model I ever used, and hoping this replaces it ;)
and? how do u feel about the new model 3 days later
In some ways better than Endgame, in some ways is not.
First the good things, this model has more flexibility than Endgame. It knows more. Really great job there.
Lightning is really impressive, I'll say it's an improvement over Endgame.
Things that are not that good....
The outputs are overall noticeably softer than Endgame and the skin detail is significantly worse. (compared to Endgame).
Lora fidelity is a little better in Endgame too.
All in all, still a great checkpoint but you have to make some tradeoffs. So take that in mind.
Anways, It's always great to see Coyotte back in action though. Great work as always and I hope he keeps moving forward Lustify series which is an amazing work that pushes the envelope of what is feasible in SDXL.
Amazing model bro, keep it up!
Version 'One Last Time' has lost a bit of hyperrealism compared to End Game due to the integrated synthetic data. I noticed a lot of my generations have a slightly more synthetic look.
@coyotte In case you do one more Lustify would it be possible to leave out the data that produces the softening and create a lora with it? This way these concepts could still be used without having an impact on the whole checkpoint.
Not sure if it is the synthestic data. I did some loras with SD generated images, but they produce sharper results than loras with lower quality real photos. However, these SD images had a lot of skin detail and were not soft at all.
@SubtleShader nah there won't be any new versions. I've tried improving the checkpoint with "OLT", but so far it seems like my questionable decisions made it only worse for a lot of people. All over again (V3 I'm looking at you). "Lustify Endgame" is as good as it can get, when you're just a dude with a single 3090ti and not a cluster of like H100's to train with. So yeah.
I'm sure you and other people will figure out how to take the best out of the version and apply that to your checkpoints. New NatVis should come out this month I think, might be useful to merge OLT with it. But that's up to you guys now.
@coyotte Im not saying its a bad thing, a lot of people here actually prefer the synthetic look for which I will never understand. For me, I want to see skin blemishes, peach fuzz, and skin pores.
@fox23vang226 yeah I understand and can see why. It's just this version was made with a different purpose in mind this time. For believable real-looking images there are V5 and V4.
V6 was more like an experiment and I still recommend people to use V5 instead.
Also, "cgi" in negatives helps a lot with that "fake synthetic" look
@coyotte Thanks for all your efforts! I understand that you had different intentions. I will try to merge OLT in my Big Love checkpoint while keeping the skin texture.
@coyotte Actually OLT does fine when replacing Endgame in my Big Love merge. No increase in softness. Seems to improve image composition and lighting.
you should have just improved fucking faces on V4, and made slightly better SFW, and it would been perfect.
"have just improved"
Man I wish it was that simple. Everything I did was to improve certain aspects of the model. The thing's that if you don't know what you're doing (which I don't ☠️), it's hard to control what gets learned. So for example by trying to fix faces with training on more portraits, you can ruin the backgrounds, etc.
Again, I DON'T know what I'm doing and this model works on black magic. Please don't expect Juggernaut-level results from a solo person without big hardware, zero sponsoring and very basic understanding of the process.
that has been a little rude.
Yeah, porn xl models have only one downside to be best: compability with loras and names.
Otherwise, this outshines every general models, especially pig-eyed, tasteless Juggernaut.
@coyotte ignore haters like this guy... incredible new model, and thanks for your hard work.
@coyotte ok sorry, If I was rude. For me V4 best model for porn. If you did this without sponsors, I can only imagine what you could do big hardware:) Why don't you open patreon or smth so that we all could all sponsor you?:)
@coyotte and Juggernaut realism is shit compared Lustify
@coyotte If you don't know, no one knows 😁. I manged to keep face style when rearranging my merge parameters, but must be 10 times more difficult with training.
to me ENDGAME works better, with my samples, and with Lora, OLT incorrectly reproduces faces from Lora.
I think this model has a different approach to endgame, they're both good, it's just that they don't have the same philosophy. Personally, I also prefer endgame because it produces more ‘raw’ photo content, not in the nsfw sense, but in the camera sensor sense.
don't listen to them! do they need realism? they just forgot how to create! keep going bro, it's great.
sorry for my google translate XD
Спасибо!
Любое мнение имеет место быть. Другое дело, что я изначально знал, что многим не понравится эта версия, и это нормально! Сейчас у меня не было цели сделать версию для всех 💁🏻
I must have downloaded about 2TB in models already, when I read endgame I made a joke, but after using it for a few days I can say that it is the most realistic model I have ever tested
"One Last Time" is pretty great!
On a second thought, now I feel OLT it's more like a side-grade.
While it's true that the outputs are softer repercuting in skin detail (there are ways to work around that with loras) and lora fidelity has some room for improvement, methinks the lightning is among the best i've seen (in a sdxl checkpoint). A lot more natural and with interesting intrincacies and nuances you don't see in an ordinary checkpoint. The color palette seems a lot richer than most checkpoints or maybe i'm just biased and i like the outputs a lot more. It sort of makes it for me. My recommendation, al always, is try it by yourself and see if it works for you.
Very interesting checkpoint. It sort of complements Endgame in a strange and unexpected way.
DMD2 variant needed. I tried with the SDXL lighting Lora and it just looks like a mess.
Download DMD2 from here: https://huggingface.co/tianweiy/DMD2/blob/main/dmd2_sdxl_4step_lora_fp16.safetensors
Move it to the models/lora subfolder, then add this to the prompt:
<lora:dmd2_sdxl_4step_lora_fp16:1.0>
With ComfyUI you need to add a Load Lora node.
@SubtleShader Just 4 steps?
@Yushio I"m.using 8 steps in txt2img as it usually gives better results. And 8 to 12 steps in img2img for more sharpness.
Now that "Lustify" is finished for good, I'm gonna take a short break and then try out what I can do with SD 3.5/Flux. This won't be anything crazy tho, cause I'm not ready to go down that rabbit hole again ☠️ Probably just some loras. So, if you have any requests - let me know in the replies to this comment.
SD3.5 is what I mainly use so excited to see what you create!
Any kind of lora that makes decent SD3.5 nsfw images xD Flux is a lost cause
@DonutsDelivery Maybe Flux gets better NSFW capabilities with the announced Juggernaut checkpoint with NSFW support. With some loras Flux already gets better now, like you can see in my examples here https://civitai.com/images/61025817 and here https://civitai.com/images/61720802. But it's still hard creating acceptable results and only soft NSFW is possible.
@DonutsDelivery If the Pony team can do what they did with SDXL, the nsfw will be fully usable for Flux. But it seemed to me that they were more focused on SD3 or 3.5.
there is no hardwired nsfw block in flux. i made a quick simple pen-is lora and it works ok. sometimes it makes fingers instead, but it's down too poor training. :)
And Flux is generally better with physical interactions, so maybe. However flux is completely incapable of making things (people and especially cloths) wet, dirty and so on. Just wet cloths, easy thing with most sdxl models, flux can not do it.
@andyx1976 I think the issue is that it is not so much a base model. it is already distilled. We need a heavily trained NSFW finetune to make it work maybe. juggernaut might be that?
Hi,
Really good job ! Appreciate inpainting model wish
Definitely a FLUX NSFW once you feel up for it.
Or some sort of FLUX LoRA. You've already proven your mastery of AI NSFW injection with SDXL, you can certainly create the best FLUX Finetune or LoRA that supercedes anything else out there.
Love your work, I look forward to see what you make next.
Just throwing some thanks your way for all the work you've done to advance realistic image gen.
Love your work man! Never give up!
Tried Flux again after 3 months. It is only doing boring NSFW stuff. Lustify is so much better and more photorealistic. I think doing Lustify loras would be a better Investment of your time.
I think a flux nsfw would be nice. The current finetunes of flux really suck, and there are already tons of loras for flux - we don't really need more, but a strong versatile finetune would be amazing!
@concedo I think Flux needs to be retrained from scratch to be as flexible as Lustify. Finetunes won't cut it.
@Zomboy finetuning flux is such a pain. It requires a cluster of professional grade GPUs, which I don't have 😅 That's why we don't get that many proper flux finetunes. But yeah, I agree with @concedo that there are too many loras already, so this wouldn't make much sense to hop on that train now.
I've been thinking about that for a long time. So it's either SD 3.5, OR:
@SubtleShader The longer I play with it, the more I realise that loras for WAN might be really interesting. Maybe hunyanvideo as well, but can't tell as haven't touched it at all. Need to find a proper guide on that matter.
@coyotte Yep, Wan is great.
I highly disagree with your opinion of v3. At first I didn't like it, because it seemed everything was too dark, but it quickly became my favorite as it seemed to be much more accurate with likenesses of loras and more consistent, less glitches, people clipping through objects I saw in some other versions. Still need to try v6 and I'd give them all another test overall, but I'm team v3. You should keep it up.
Dude, you knocked it out of the park with this one. The women I've been able to create are stunningly beautiful. Thank you brother.
Thank you!
Hi, thx for the work.
Is there inpainting model coming soon ?
appreciate
I can try. Do you think the inpainting version should be based on V6 (olt) or V5 (endgame)?
@coyotte V6 imo, really appreciate, great work :)
@coyotte i'd say V5 would be much better.
OLT version seems very polished and I guess that's because of synthetic images in its dataset, which usually gives squeaky clean and polished (PONY style images). this is not bad at all but leaning towards realism would always be appreciated and used by many.
WOAH, this last one is something else. Adherence with dmd2 is next level. Very different to your last models. IMO, you're onto something big with this one and can't stop here. This is another beast all together
Thank you!
Honestly, I feel like I did everything I could do and could think of.
What's lacking right now can be improved with loras, which will be taken care of by community if this model becomes the standard. If not, it will mean that there is something better with a wider concept knowledge, and that's fine.
@coyotte Just tell us your "familly account" name to allow us to know what are the SFW models you are cooking or preparing !
@Brewce haha, there's no such account 😂 this is the only one i have, everything I do will end up there
@coyotte Hi, what is the difference between normal and DMD2? I look forward to your answers (I just discovered this model, it's pretty cool) PD: And I was wondering, does it work well in inpaint or is there any other model that is almost perfect in inpaint? Thank you very much in advance!
@Eliz103 Hello!
So in short, DMD2 verison's purpose is the same as Lightning/Hyper would have - to output images using as low as 4 steps instead of 25-30 as a normal checkpoint would. This comes with tradeoffs, of course.
You can use inpainting version of lustify to inpaint, there is one in the versions tab
@coyotte Wow I understand, and in your words what would those tradeoffs be? About the inpainting model, yes I saw it late and I'm using it just it's pretty good but I realized that in FORGE I can not use the integrated controlnet, is there any way to know why? (Without controlnet is also very good) (And in A1111 controlnet extension works) I hope your answers, thank you very much indeed! ✌️😁 ( RuntimeError: Sizes of tensors must match except in dimension 2. Expected size 1000 but got size 1040 for tensor number 1 in the list. )
I have copied the workflow from your images and when I search for DPM++ it always generates glitches in the images.
Um, there's no specific workflow in most of the example images. These all were made with basic text-to-image from Forge/A1111, so I'm not really sure what exactly you have copied.
Also, it should be DPM++ 2M SDE, not just DPM++.
I'd recommend you to start over and use something basic. DPM 2M SDE + Karras/Exponential, 25-40 steps, 896x1152 resolution.
I'm having problem with this model when merging it with others and using ADetailer. I'm getting washed off colors and strange white dots when mask detection, then it all goes to hell. Is it because of VAE?
Huh, interesting. It does sound like a VAE problem, but I'm not sure why it appears only when using ADetailer. What happens when you try forcing a vae?
There is one baked into the model. It shouldn't happen.
Did you merge it in Forge? That feature in Forge is broken. Either use A1111 or ComfyUI for merging.
It feels like a step backwards from EndGame. (IMO v2 was the all time best). I tried OLT with a simple 7 word prompt that v2 or any checkpoint would do in it's sleep (beautiful naked big breasts Asian girl masturbating) with zero LoRas to get the best idea of it's default output, and a run of 6 images on
I'm using Fooocus, so Fooocus Enhance is OFF, and just the default styles are on. Speed setting, everything Fooocus related is set to default.
It has a habit putting intense lights shining out of nipples/vaginas/mouths/foreheads, giving the girls oily skin for no reason.
If I try and use it with the same LoRas i've always uised, then it completely falls apart. Blurred faces, deformed faces, low quality output.
When it does turn out something of acceptable quality, it is like a 1970's polaroid. Which is not a style I set, or asked for in the prompt.
I think I will go back to v2.
Thx for your comment, gonna try it, because unlike Endgame DMD2 i almost got the same results as yours while testing several versions so i sticked with endgame. Its better V2 Lightning or the default one ?
@NinehundredBr Maybe because v2 was the first one I ever used when I was a total n00b to this stuff I am remembering it with rose tinted spectacles like you do your first girlfriend. Mostly I just use Juggernaut XL these days, and only use Lustify if I need to have a dick in a picture, because it does dicks much better than Jugg. I'm about to download this new fixed version and see how it is anyway
does this need a vae? all my generations look pretty bad, like if they were missing a vae
No VAE, no image. It is baked into the model. You are doing something else wrong. Maybe wrong sampler, too few steps etc.
I'm always having the same error, how can i find the right vae for this?
Given groups=1, weight of size [512, 16, 3, 3], expected input[1, 4, 128, 32] to have 16 channels, but got 4 channels instead
A proper one is already baked into the model. Don't force load VAE
jpge
update inpainting SDXL version pleassssse
Yes, I will when I have a chance
@coyotte much appreciated king
i second the request @coyotte
@crombobular @napsumfur225 what UI are you using guys?
@coyotte i'm using reforge. i have done some inpainting with endgame and onelasttime, but assumed they weren't as good at it as an actual inpainting model
@crombobular okay, I've asked in case you're on ComfyUI to share the workflow I've made and had very good results with for inpainting. It uses fooocus's inpainting modules imported into comfy + lustify.
Inpainting model will be uploaded later.
Done.
@coyotte Can I have the workflow you made for comfyUI?
just discovered the lustify models. what's the main difference in v6? you say it's different to v5 but how?
Version 6, unlike 5, was trained using also synthetic images generated with AI, so it is capable (in theory) of generating less conceptually realistic images, while still maintaining realistic quality.
To give you an extreme example, if with v5 you were not able to generate a woman with the appearance of a fire-breathing dragon (for example), with v6 you might be able to generate it.
@BobDF thanks a lot for replying to them! I was going to do that, but didn't have enough time to write this all down
It also was trained for additional massive 800 000 steps. Earlier versions hadn't had that much training between them.
I really wish people would include prompts. Because some of these photos look too good to be ai.
You can filter so that only images with prompts show up. I don’t look at an image if no prompt; what’s the point?
too good to be ai? welcome to sdxl (what it could've been right from the start). some people don't wanna share (rightfully so), get over it. isn't the image alone enough? try and recreate what you like, might even learn a thing or two.
@jmueller I make great images. So don't patronize me. Asshole. And yeah, you should share prompts. Because people do share non ai images here. Shoes how much you know 😂
This person https://civitai.com/models/1351339/thebloke-realistic-xl literally stole my model (Lustify Endgame) and claims that it's his own. I've made a comparison on 3 different seeds and prompts and it's indeed my model renamed. They also hide comments and pictures that point to that fact.
After contacting them, they block me instantly. I've already messaged mods, but your help is needed as well. PLEASE REPORT THIS SCUM.
Proof:
https://civitai.com/images/64757087
Done :)
this is sad. They don't even merge it with something else to try to hide it. Very low effort.
it's wild that civitai just let him go even though he's stolen other shit too.
hey dont worry will donate something testing V6 also why dont you put it behind a paywall or something like patreon so your work is not lost
Done. If he stole yours he's likely stolen everything he's posted, why put any effort in when he can coast?
@crombobular the funny (or sad, depends) thing is that Civit doesn't really punish people like him. At least in my experience. They delete the stolen model and let them be. I've already encountered that, just didn't make it public. So they keep going around and stealing stuff.
@sonic4life170 Goes against my principles. I'll never paywall my stuff, neither I'll make it "early access" BS like a lot of people there do.
@admiral_underpants Thank you. Yeah, I'm sure everything they've uploaded is stolen from someone. I mean, their account was made 2 weeks ago. That just screams "SCAM".
@coyotte yeah i understand its lame the internet is full with idiots and piracy i understand maybe they can block them i hope so
The 'Free AI Program' he links to sounds totally legit too.
@coyotte looks like it has been taken down
Except the dudes on TV and is like famous for AI. Seems like someone is jealous.
@crombobular Nope; this site allows users to block individual accounts from viewing their model, so if you called them out on the model page they've just hidden your comment and blocked you.
@highfalutin holy shit really? that is really stupid.
yeah, wow. I reported him. "moderator review needed" and explained things. you are doing great work, don't let this mofo get to you.
@ElectricDreams it does appear to be merged with something else, there are subtle differences, but it seems to be about 99% Endgame.
@crombobular Still up, I think the op just changed his user name.
I've done some tests and can confirm Coyotte's findings. My proof can be found here: https://civitai.com/images/65032212
@sdxl4life We can all see your account was created a few hours ago. We know you're a cock puppet. We know who you really are.
It seems we've won the battle... for now. (sorry for spamming).
@prodajie YEAH WE WON! Thank you all for the help. And thanks to Seb (moderator) from Discord, they were really helpful.
Yeah, that guy probably merged it at 1% strength with something just to change the hash, so that Civit's automatic moderation system doesn't get triggered.
@coyotte You just need to change the metadata a bit to change the hash. Merging at 1% already produces a rather different output.
@prodajie it's really sad.
Some people are just shameless and sociopathic. I make a lot of mods for games and people rip my mods verbatim and change no code at all and just re-upload it with a different name. I seriously cant understand what goes on in these kind of peoples minds; zero sense of honor or shame.
Sorry for any spam you will see, for a minute there I thought I was shadow banned and tried posting a few comments back 2 back to test if I was or not. Turns out I think the servers are just extremely slow today.
@fox23vang226 yeah, civit's been under the weather for sure today. had to reupload my refiner workflow two times
@fox23vang226 these people probably aren't good at anything, because they never had patience to get so. they want to look otherwise in their own eyes. it's a weird unhealthy way to cope
@prodajie read his article. This dude is advertising a literal fraud
@coyotte The guy also has his own meme coin 🫣. Too much self praise. Another narcissist.
really good! does anyone use this with flux? Im using generate img Flux > Upscale with realdream(ultimate sd upcaler), what config its recommended? to fix eyes/skin of flux model
why would you wanna do that? if you want to go that route do the initial generations with lustify and use flux for upscaling - but I find that lustify does a pretty good job at upscaling & inpainting aka waste of time (imo)
The problem doing Lustify > FLUX is that skin sometimes feel off, and the nsfw can be missed
A LOT OF STUFF COMING OUT TODAY.
OLT WITH FIXED TEXTURES.
2 WORKFLOWS.
PREPARE TO TAKE YOUR PANTS OFF
i've been using OLT basically non-stop since it came out, i thought it was freaking amazing, but now you're saying there was something wrong with it? "smother textures" (smoother?). I guess I can see it now that you mention it, but I thought it was a feature, not a bug. xD
I'm not wearing any pants ;)
Great work. Looks much sharper!
@StanleyPain everlasting goon sesh, THAT'S WHAT I'M TALKING ABOUT
@shapeshifter83 I didn't see it as well, until I've tried fixing it. Then I've realized that people were right. And yeah, it should've been "smoother", not "smother" xD Thank you
@SubtleShader thank you!
PANTS WERE ALREADY OFF
@coyotte Always. Now and forever :D
@shapeshifter83 same here, freakin love OLT. Gonna test texture fix tonight
I N P A I N T I N G
@napsumfur225 I wish there were more hours in days :( I remember about inpainting version. It's just that what I was talking about was planned a long time ago.
@napsumfur225 Done.
Details
Files
lustifySDXLNSFW_oltONELASTTIME.safetensors
Mirrors
lustifySDXLNSFW_oltONELASTTIME.safetensors
lustifySDXLNSFW_oltONELASTTIME.safetensors
lustifySDXLNSFW_oltONELASTTIME.safetensors
testeone_v10.safetensors
lustifySDXLNSFW_oltONELASTTIME.safetensors
lustifySDXLNSFW_OLT.safetensors
lustifySDXLNSFW_oltONELASTTIME.safetensors
lustifySDXLNSFW_oltONELASTTIME.safetensors
lustifySDXLNSFW_oltONELASTTIME.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.