An interesting note to put in top of the description:
I've just reread all the recommendations made by nutbutter according to how BigASP2 was trained and I've done my own tests.
If you want really striking results, don't try to invent tokens, just look at this list and the tokens used for BigASP2. If you want to save time, take a look here
Take the tokens that will have the greatest impact and mix them together to obtain high-quality, personalized results.
This way you'll have fewer deformed bodies, deformed hands etc...
It is essential to understand how JoyCaption formats captions for optimized use of fine-tuned or merged models using BigASP2.
If you apply this recommendation plus the fact that you generate resolution higher than 832 + 1216 (1200 x 1552 is a good sweet spot without HiResFix) while keeping the same ratio : that's a bingo. 
-It's just a mix merge of the best models from the Lustify collection created by coyotte.
This merge contains the OLT + Endgame + Alpha V4 version.
Some people have complained that the OLT version looks too "synthetic", I'm not entirely agree, this version is in fact very good but requires a bit of time to understand and adjust the prompts.
But since I like experimenting and trying things out, I thought why not mix the best of Lustify and see what happens.
According to my tests, with models such as lustify / BigASP... You can push the resolution up to 2048px (Even if I prefer to stay in my safety zone around 1800px) without using HiresFix but by using Kohya HRFix. This allows you to gain a fair amount of definition and detail without encountering too many distortions (even if this happens more often) and also means faster generations.
Just try to start from this basic ratio of ‘832x1216’ and gradually work your way up while maintaining this ratio.
Why this ratio? Because nutbutter has used a lot of reference images with this resolution to train BigASP, so I may naively think that I have a better chance of getting good images if I use a higher resolution with the same ratio
I always use this useful plugin if I want to lock and keep my ratio.
I also strongly recommend changing the "inpaint width and height" settings in Adetailer to 1024 and the use of the "mediapipe_face_mesh" model. Obvious for some, perhaps less so for others
🔥hint : if you want to add more precision / detail to your generations I also recommend using ‘PerturbationAttentionGuidance’, don't go beyond scale 2, I usually stop at 1 (Values that are too high can increase the number of deformations.).
I also use PAG for SD upscale script. With Denoising strength around 0.30.
So basically, my settings are like this for PAG and Kohya HRfix :
For the rest, I'm mainly using DPM++ 3M SDE + Karras at around 40 or 50 steps with a CFG scale around 3 and 4.
Don't ask me technical questions, I'm not a model checkpoints creator or finetuner, I just make Lora for fun and try to understand how to push the limits of SDXL.
Making a merge doesn't even take 5 minutes so all credits go to coyotte and nutbutter
(the hardest part is finding the right merge with the right settings, know at least where you want to go, a bit like trying out a new recipe.)
Merge was done this way :
Maybe it's good, maybe it's bad, half good or bad.
I don't know, I like the results I get and that's what counts.
Try your own and see 🐈⬛
Description
Mix merge of the best models from the Lustify collection created by coyotte.
Contains OLT + Endgame + Alpha V4 version.
FAQ
Comments (19)
in early 2025 and sdxl cant do a realistic anatomy from side or side back , without doing cartoony carved lowerback , upper limb always seems like photoshoped to another persons lower limb that comes from space not human at all. sdxl don't know what "butts" look like , butt cheeks and many more.
i think you are doing something wrong, what do you use for generations?
@Gumibeer77 comfy cfg1 dmd2 lcm beta , no prompt can fix this wrong anatomy , and about me doing something wrong .. the 95% of this pose in images posted here has same issue .. all my ear ...wait your reply
@amazingbeauty Use an anatomy LORA and you're good to go!
@SantaWithAI where is 'an anatomy lora' you speaking about ?
it is really the same for sd 1.5 models, you cannot create realistic looking bodies without a super defined and visible ribcage, sometimes even when the prompt specifies for non thin characters
@the_zenithsun true , and making realistic body with recent fine tuned sdxl is possible BUT with poses things starting going unrealistic , ribcages thin boony alien shape with body pumps , then for very simple standing pose there's never exist sdxl model at this moment i type this can do standing body from side or side a bit in realistic shape it all can do that carved lower back like hell , other thing and it's anatomy error also , no butt cheeks at all , other anatomy error ribcage , other anatomy error belly botton is way very high it's kind of closer to be positioned to the breasts and that not real ..all that with standing pose , i can't make nay sh*t picture if just standing woman !!! all is crap carved aliens ..it's thin bone or fat ass ballon , real body shapes not like that at all , when you try make regular body it all gone fat one piece . realistic body not like that , sdxl and all image generation models still suck ass (2025).
@amazingbeauty Dude, how many samplers and models did you try? just because your settings with one model does not give you results you want, it doesn't mean that all SDXL sucks, i would say that you need to improve a little bit on your knowledge and skills. i can see that already on newest version of this model that bodies are more realistic. good luck.
Bodies are completely realistic, you just have to avoid betting too much on tokens oriented to get the "perfect look", I even sometimes put score_1, score_2 ... or bad photo, or poor quality in the positive prompt, it gives me quite realistic and more amateur / candid results. In fact, you just need to understand how BigAsp, especially 2, was trained and how joy-caption-alpha-two allocated the captions/tokens to this model. I'm myself currently training a LoRA with 5500 images captioned with joy-caption-alpha-two. If the result is convincing, firstly I'll make the LoRA available, and secondly I'll try a merge with it. Its mainly focused on natural photos rendering (SFW and NSFW) and much less selfies (which tend to deform bodies because of the wide angle). Just a note to finish, merged or finetuned models with BigASP perform better at higher resolutions, since BigASP has been trained to perform beyond 1024 and I do exactly the same thing with the Lora I train (mainly with bucket resolution of 960 x1472 = count: 7876 / 1472 x 960 = count: 2566).
@Gumibeer77 one model ? sorry i will not answer you because your imagination is low
@Gumibeer77 but just to be real before i leave your reply , try ti prompt woman from side or side back , and from back , get those 3 pictures and compare them for same pose of real pictures of real women or girl from internet.
"just use an anatomy lora bro"
can't spend 20 seconds to write the name of one
@chrmaroudas625 I am going to use a lora to render them giving birth
Really appreciate the honesty and the reasoning, and thanks for the guide too! Cheers~
What's the different between LustiMix & BigLustiMix?
BigLustiMix is a merge between LustiMix and BigLust 1.6. With the aim of recover a slightly more "photographic" aspect in the raw sense of the term. BigLust has a "muted colour/flat" amateurish look. I like this side. And above all I wanted to keep the latest Lustify improvements with endgame and OLT. So it's a kind of BigLust on steroids
@WearCatEars which is 'amateurish' from your models , please write the full model name ....??
please ...use unique names
The first merge is the main mix (LustiMix)
the 2nd merge is LustiMix with BigLust1.6 which itself is a merge of 60% BigASP 2.0 and 40% lustify v4.
The word ‘big’ comes from BigLust, which comes from bigASP, and is a logical sequence of model names to help you understand which models they belong to, so I called it BigLustiMix.
It's not really difficult to understand, I've described everything in the "about this version" section, just read.
Since I like to have a choice, I release both versions, there's no one better than the other. I'm not going to put LustiMix v2 on, as I don't see it as an improvement, but more as a complement / alternative.