As a rank beginner with maybe 16 hours experience from "starter models" and suggested stuff, this is far and away the best thing I've tried.

I've downloaded stuff that does some of the things this does, but they interact weirdly with each other or require incompatible settings and models.

This does it. Sure, it hates wearing pants. Ain't that what we're here for, though?

I'm a big fan of your work. Thank you for the continued updates.

its really great but sdxl is clearly showing his limits

Yes I tested the latest version 7 , and I must say the quality is much better more realistically, also shadows , and more details, also the women are hotter 😉 🔥 more sexy, and the backgrounds are more detailed. That's from my experience, and I cant wait till it even gets more realistically, but I dont know if that is possible with sdxl, I only use comfyui and I dont know how to train, but I enjoy this model. So thanks

fans and haters both are demanding an inpainting version of GGWP, fans would like it, haters would appreciate it, and both will be happy at the end of the day :D

anyone know the best settings to train a consistent character lora on this model?

Please train this on the Chroma model i think it would excel this !

uhhm, i created some really nice images with lustify, but as of late i really get annoyed by the anatomical flaws. some female hands and feet just look scary. a real turn-off for using this model

Love it! Any tips on making futas? It keeps coming out with the same face and the face is kind of manly.

I'm still using Automatic1111, and the upgrade from 6 to 7 using the same prompts that I was using successfully in 6 to create great images are giving me super low-quality images. What am I doing wrong? Or anything I need to update like VAE? Or is this an A1111 thing?

Can anyone explain what DMD2 means here? i just see the file but i have no idea where to put it, is it textual inversion? is it something else? Using stability matrix forge ui. Please any help would be appreciated. Thank you

V7 is epic! amazing results! its my goto model! can you please make a V7 inpaint so it will match quality for little fixes?
Thank you so much!

Inpaint ver. pleas

I love the model. I have melted my GPU finding a decent model for SDXL since pony doesnt do it for me, v7 might be the best in the website. In case the creator reads this, i would like to ask, any way of having more control on futas? namely body shape and d size, If its not baked in the captioning, will v8 have these in? Much love coyotte:)

for amateur pictures this is extremely good.

Love this model. It's great and is extremely powerful. But somehow for me the pictures have a palish hue. And also the face sometimes looks kinda "plasticy". Maybe it's only one me though. That's why i switch up between the new and older model, depending on what i'm doing. But it's really good at creating amateurish images!

Any chance to train the model for more advanced cards -- like 16 GB or 24 GB? Superb work. Excellent model.

Hey @coyotte , is it true you were aroused the whole time while gathering the dataset for this great model?

When is ver 8 gonna come out?
and can you make it render better and handle loras better like Juggernaut XL?
please and thank you :D

I"M fucking back to make money hehe thanks bro I love you I already gave you 200 bucks so dont cry

great for generating stunning women and simple porn. if you want more complex images lustify gets really freaky and doesnt understand any prompt and creates ugly freakshows.

Great model as to be expected but I find it has a major problem with nationalities and ethnicities. Even some pretty common ones are just impossible to get, I wonder if they were even trained. A real shame. Otherwise great job with this one.

Hopefully v8 will be better at this.

How to close mouth to all characters in this model?

Bro, I'm totally new to this and the google drive link is not working, can you please reshare it? :)

If you like anal and trannies this is def the #1 checkpoint. Gl getting a d1ck inside of anything but her a-hole. Bonus, if u keep trying she'll just grow a dick.

Looks like google removed the DMD2 workflow, could someone or the creator reupload it? I've been using NAG with moderate success, but hasn't been reliable.

Please reshare the DMD2 workflow Google took it down, as well as the DMD2 you use too. Thanks!

can you resend the link or google removed it

i beg - please make a GGWP V7 Inpainting Version.

In the DMD2 workflow the last stage of the process is tile upscaling, but the generation is saved before this, and the tile upscale result never gets saved? Is this meant to be the case? Maybe someone can give me some insight. Thanks

Tried using both the DMD2 and non version of your workflows and I'm getting the following error:

Error

No link found in parent graph for id [4] slot [0] clip

I see no errors in the console and none of the nodes is outlined in red, so I have no idea what is missing.

Jumping from one error to the next in your WF.
KSamplerWithNAG

NAGCFGGuider.inner_sample() got an unexpected keyword argument 'latent_shapes'
Any idea?

Do you have safe default settings for V7 using the DMD2 Lora? I don't have Comfy so I'm not able to see what you have on the workflow.

Currently doing LCM Exponential, 8 steps. HiRes Fix 1.5x, 0.35 denoise, 4 steps.

Results are good at these settings, but wanted to make sure I wasn't missing something.

InstantId doesn't seem to be working fine in the v7, tried changing the settings several times but while the face stays consistent, it destroys the prompt. Any recommendation?

I really like the previous Lustify checkpoints, but with v7 I am not getting any good results at all. It only works well with the DMD2 lora, but with "normal" settings, (DPM2, Karras) I am not able to get good output at all, and on same prompts that worked well on earlier checkpoints and that produced similar image compositions I am getting a completely different composition with v7. Can someone help me, I would really like to access the greatness of v7!

I think the anatomy and overall look is good.. bit better than last version, but main problem for me is that every image seems to put a dick in it, either futa or a man even with very strong negative prompts against - worse than previous version. It would be great to see a NSFW model with more solo and lesbian capability / training. If anyone has any tips on how to get rid of the dicks I'd really appreciate it!

Is there any way to prevent all generations from having tan lines? I am using the inpainting version and v6.

is there anything that differ when training a LoRA with this model(realistic in general) over anime models(e.g WAI)?

I can't get any LoRA's trained here to work as good as i can with anime models

Since the release of V7 I have trained 56 loras of various type using it as the base model. Not only does it make the crispiest most accurate Loras of any checkpoint I have tried, but when combing them back with V7 it is hands down the most realisitic images i have seen from any model, Flux, Qwen, Wan, Pffffffffft! .... Now finetuning the 36,000+ img dataset. Initial tests are indistinguishable from real life. This is the only checkpoint I will ever need now, Coyote, youare the checkpoint fkn masterrrrr! -- I long thought SDXL had come as far as it ever could until you dropped this abso-fkn-looot banger! How was this possible?

Works great as a refiner for my pony checkpoints. Takes the plastic face away and my loras faces look now like the source! Great work!

Adding extra penis to the women 😂 I could not find any solution for that

Is this normal? https://imgur.com/cHQhhnc

I don't know what i'm doing wrong, i just put the checkpoint at DMD2 with Leakcore lora. I get that as a result. Sorry if it's simple to fix, i'm still new with ComfyUI.

What are the recommended settings for the adetailer?

I'm new to LUSTIFY and have a few questions: 1. Does this model understand concepts beyond the common ones (missionary, doggy style, cowgirl)? Is there a list of tags I can use with this model? 2. Can all the style tags (1 through 7) be used in the same prompt, or do I have to choose just one?

DMD2 version works great, but v7 has some problems on face details

how are yall using this my comfyui nag will not download when trying use dmd2 it gives me an error message

For DMD2, what is everyone using for LCM scheduler? Hires fix settings? SD Ultimate Upscale? For Hires fix if I run LCM ddim_uniform it's extremely sensitive to steps. It's most stable at 2 steps, and low denoise of about 0.10. If I try 3+ steps it doesn't matter how low the desnoise is it will hallucinate anatomy severely.

How do you use the inpainting portion of this workflow?

This doesn't even show up to choose in Fooocus and in SD it produces complete garbage images. No idea how anyone got this to work at all.

A model made by the gods

Awesome model and workflows, thanks a lot!

What is the difference between the DMD2 workflow? I'm comparing both with same seed and prompt, and I can see there is a difference, but wondering if someone can explain the benefit of each. Is it only speed and personal aesthetic preference?

Will a V7 DMD2 version be launched in near future ??

Did anyone launch thit model with 4 gb vram? 😅

thank you

what is the best resolution ? basic question but i'm not really sure

er what is DMD2? google is not my friend on this one

You really killed it on this latest version, great diversity and detailed backgrounds.

"and I do a lot of dumb-crazy stuff that in theory shouldn't work."... That is some pretty awesome self-deprecating awareness. I tip my hat to you.

You you plan to release a ZIT version?

Couple of questions:
1. is there a way to generate two or more characters in the same image (with different features)?
2. Is there a way to choose the breast size and keep generating the same character with diffeent poses/situations? (everytime I change something it generates a totally new character).

@coyotte Regarding character LoRAs with these checkpoints - is it better to train them on SDXL base model or choose my favourite Lustify checkpoint and use that as the base model to train the LoRA on?

Hi, thank you for cool modes, the links to workflow in description doesn't work. I wanted to ask if I can use OLT inpainting as in workflow in images, or there's some updated version?

the link is not working

Any plans for Flux.2 Klein? 9b base model looks promising

Anyone can give workflow json for the image edit?

2 question:

This version is better than v4 alpha?

BTW, im getting this error when i try to use the non-dmd workflow

Cannot read properties of undefined (reading 'workflow')

Hi, I have a question, using SD with A1111 (web ui 10.2.1) on Runpod I have perfect results, but when it's on local (newest version) the eyes are always off, melting or weird. Is that normal ? The Checkpoint can't handle the newest version of SD ? TIA for your help

I was skeptical at first, with this one, couldn get any good results, but now I agree, this model is truly a masterpiece!

one of my favorites. just wish it had more..deepthroat 😅

Hi. I'm getting the next error:
InpaintModelConditioning

GET was unable to find an engine to execute this computation
I'm using ComfyUI Zluda with an AMD card.

for anyone new to this page, the model you actually probably want is to the far right on the list - the original OLT.

How do I get the upscale working. Everything else works great. As soon as i turn on tile upscale it wont run. I have the models loaded. Haven't used the inpainting yet tho either.

How do you move from a generated image into video with this model? I assume using WAN 2.2. Any step by step or pre built workflows?

This model is okay for doing amateur photos, but if you want hard NSFW you'll need loras. Idk why, but this model can't do c*m properly.

The greatest checkpoint ever made. Been using it since first release 70,000+ images - then V7 took photo realistic GOATism, the results I get MOG every model.

Only now with the Z-Image release has LV7 not been a daily feature, However I still keep coming back as there are certain images that Lustify is more realisitic.

The dream is Lustify Z-Image.

There was no other choice than to painstankingly go through the arhcives, collected 20k of the most photo realistic Lustify images from our collection

As we speak our RTX Pro Workstation is hammering anunofficial Lustify Z-Image Base. Hoping and praying eith every epoch"

Which app should I use to generate videos with the v7? I used Fooocus with v5 endgame checkpoint, worked ok, but now I want to try out videos.

I dragged the json into my comfyUI workflow, it prompted me to download 8 missing requirements.

However, when I try to run the prompt I get this error:

SyntaxError: Unexpected non-whitespace character after JSON at position 4 (line 1 column 5)

New to comfyUI, can anyone help? Thank you!

This is the best inpainting checkpoint that worked for my needs.. Thanks my guy! Other ones were really bad, or just didn't work at all, but this one gives such natural results with very little work.. I made an AI create a workflow JSON put the model in and it just works. Fucking brilliant man!

waaaay too many weird deformations, too bad

this is the greatest checkpoint ever! on a regular basis I am checking for a new update or see what the creator is up to.

I have tried so many, none come close to this one.

I have filled a 2 tb hard drive just with sdxl checkpoints. if I could keep one, its this one

Hey @coyotte ! First off thanks for all the hard work you've put into this. It's gotten me into a new hobby that's been lots of fun to explore.

One question that's been nagging me for a little bit is why, in the NAG/DMD2 workflow you created, does my workflow error out when I include an ipadapter facID after the LoRAs and into AE? After trying to troubleshoot with Gemini and GPT, I've just determined that I had had to choose one over the other (which is a bummer, because otherwise I have things running really well!) Are they trying to sit at the same seat of the table sort of thing? Is there anything I could try and do to get them to work together in the same workflow?

Lastly, should I stick with GGUF or utilize OLT? Do you have any opinion as to another experienced users advice in including a contrast node at negative to run a higher cfg or DMD2 LoRA weight? Thoughts on using fp16 AND fp8 DMD2 at lower weights?

I know this is a lot of silly questions for a newb like myself, but any advice would be super helpful from the man himself. Thank you!

Had some trouble making your workflow work with this. Especially since It was my first time on ComfyUI but goddamn the results are insane. Freaking love it. Just wish it was slightly easier to create SFW in the same style it got a tendency to really REALLY want to show boobs lmao

I tried to use the DMD2 workflow, but I do not get it to work. Installed missing nodes etc. But directly after hitting run i get:
Error

Cannot read properties of undefined (reading 'output')

It is almost a year passed since OLT was released. It will stay in AI history for me as greatest SDXL checkpoint of all time, true original one. And it still deliver exceptional quality with proper settings.

Just want to give my respect to @coyotte for it and LUSTIFY! series in general. True legend.

Hey. Thank you for your work! I tried the NON-DMD2-Workflow. But I get an error message: "can't access property "output", res is undefined" Do you have an idea how to fix this?

Hey, thanks for the model. Big fan.

Anyway, did you use the workflow recently in Comfy? I'm getting errors like other folk and I'm wondering if it's maybe using some nodes that weren't update to the new Comfy nodes? Any ideas?

Hey bro! Is it possible to train a LORA character at your checkpoint?

Greatest model of all

I now check or the 300th time to see if there is a new version

I'm novice in this matter, so I'm sorry possible stupid question, in description here written that this model isn't ILLUSTRIOUS-BASED, but in 147f5i recommended workflow used illustrious checkpoint and illustrious loras why? if illustrious shouldn't be used, what should be used?

gooner nerd full send! my wishlist for the final version of lustify:

- better handling of sheer/see-through fabrics and/or mesh/fishnet

- a re-injection of some original SDXL knowledge or new knowledge of persons/faces (i know that's taboo and can't be acknowledged even if it happens behind the scenes)

- better nipples, more prompt obedience with nipple variations (long_nipples, etc) and nipple-related tags (visible covered_nipples bumps aka "pokies")

- better prompt adherence with crotch-related anatomy or scene (cleft_of_venus / innie, long_labia, deep_penetration, etc)

- pov prone_bone knowledge (legs_together adherence)

- an improvement of general visual quality from GGWP

- makeup/hair related prompt adherence

How to use with dmd2lora cfg sampler scheduler? Steps etc..

I want to create real girls , lustify is okay but the faces are sometimes like they need the yellow bus, how can i make high resolution faces with fscedetailer, what are the best settings

For the next version i would love to see, higher resolution faces, and mouths , detailed skin texture, no missing hands or bad fingers, no deformed eyes, more natural girls, curvy, skinny, better lighting, shadows, real looking women, chicken wings, 4 fries, cola . I might donate with bitcoin again but lets see some new version

One more question please, why if I use cfg more than 2.5 in this workflow tjya84.json colors of the image become almost poisonous bright?

does training this kind of content desensitize you? I'm picturing the hours upon hours of analyzing and captioning the training images. I would probably need long breaks between model training to make my wee wee work again.

where do i download the workflow? the link isnt working for me

Will a standard SD XL trained lora work with this? Or do I need to train a lora directly on this checkpoint, is that even possible?

You did it. Son of a bitch, you did it! 😁

Just got v 8 through your boosty and your workflow. The quality is insane, I think I found my new favourite model - the sharpness and textures are even better than in OLT! Thank you for your work, I especially liked the fact that 1536 resolution is now supported. This is the first SDXL checkpoint to achieve that, I think?

TLDR: it 100% worth it's money.