Krea 2 is an incredible image model, and it's looking more and more like it will be the true successor to Stable Diffusion 1.5/SDXL in terms of mass adoption and post-release training (LoRA's, model finetunes, etc). But the base model and all the other LoRA's (so far) have only focused on the, uh.... roast beef sandwich aspect of the vagina. Lots of wrinkly meat folds (very pronounced inner vulvas). And if that's your thing, I got no beef with you.
But for me, I like a cleaner look. Smooth curvy lines that are proportionally plumped up to an aesthetically pleasing degree. Simple, basic, not complex.
---------------------------------------
VERSION 2 UPDATE: So far the #1 complaint about version 1 was that it created extremely smooth skin that lacked details (very plastic/3D looking). I have to admit, I had tunnel-vision when testing that LoRA and actually thought it looked fine. It's possible I did the same here, so you be sure to let me know so I can keep fine-tuning for future updates. I appreciate any feedback, even if it is negative, harsh, or just poking fun at how bad the LoRA is, lol.
Anywho, I completely overhauled the original dataset images by meticulously feeding every single one of them through an AI image editor (I used Klein 9B KV with a few good skin-texture/realism detailing LoRA's). I dialed-in each image as best I could (and to match the style of the original as best as possible) and trained at rank 32 (instead of rank 8). I trained to 3,000 steps like V1, but I actually found the 2,000 step version was much better (probably less overfitting). And that is the one I am providing here. I may train well past 3,000 to see if anything improves, and if it does, I'll do a 2.x version update.
The quality difference between V1 and V2 is night as day in terms of skin detail, but it's still not perfect. I may have gone a little too far on the skin texture detailing, as things can sometimes look a bit wrinkly or dried out (or like the character has some kind of skin condition). But hey, at least it's better than smooth non-detailed skin, right? Plus you can always adjust it down or up or mess with the samplers/schedulers to get exactly what you want. Euler/simple is the most balanced and euler/beta amps up the skin details. And if you do 12 steps on the sampler (instead of 8), it will add even more skin textures/details (usually too much, but you be the judge of that).
So while V2 is a huge improvement on realism/skin detail compared to V1, it is still not at the level of Krea 2 base quality. Maybe that is impossible, I don't know. But I did a lot of A/B testing with my V2 LoRA and the Krea 2 base, and the skin detail/realism was decently close (at least to my eyes). But my LoRA definitely does reduce the level of overall detail compared a baseline Krea 2 sample. Clothes might not be as detailed, there will be less stuff in the backgrounds etc.
Give it a try and let me know what you think. I've only done a very limited amount of testing (a few hundred images/prompts), so no doubt I am unaware of a lot of potential downsides (or maybe even some unknown upsides). Either way, for those who really hated that V1 was so smooth and lacked skin details, this should get a little bit closer to what you may want.
--------------------------------------------------
m99's Labiaplasty LoRA's are, IMO, the golden standard for innies, and I hope they eventually release something for Krea 2. But since training this model is surprisingly fast (and very effective), I thought I would give it a try myself. And I'm honestly shocked at how well this LoRA turned out (at least with the tests I've done so far).
Training Info:
My training dataset consisted entirely of synthetic images of women that I generated using m99's Labiaplasty LoRA with Stable Diffusion 1.5 (Cyberrealistic) and Flux.2 Klein 9B models. On all the images only the lower body (crotch/butt) were cropped/used, so there should be little (ideally none) influence on character faces/hair/body or even overall model styles or compositions.
This should stack with other LoRA's very well, though it will be diluted and will compete when stacking (especially if the other LoRA is also trained on vaginas), so adjust the strength of it and others accordingly to achieve a proper innie. From what I've tried with stacking, it almost always has a positive effect on the vagina even if it doesnt make it a true innie (plumping up the outer lips and reducing the overall inner vulva size). It's actually quite nice and allows for some great vaginal customization in this regard when combining with other vagina LoRA's or all-in-ones.
Usage:
I did not add a trigger word or any captions to this LoRA. To make it work, just enable and set the strength to whatever works best for you. It is trained at 1.00 strength. If stacking, you may need to adjust it higher (or adjust other LoRA's lower) to get the right results.
Description
7/1/26 - Version 1 known issues: While this LoRA does work for its intended purpose (and pretty decently), it has the unfortunate downside of really crushing skin details and making the characters look more 3D/plastic/smooth. I'm already working on a Version 2 which will hopefully bring the level of realism/detail closer to the base Krea 2 experience. But until I get that finished, in my post-release testing I've found using the dpmpp_sde sampler with the beta scheduler does seem to help a bit, but it does take longer to generate (about twice as slow as euler/euler_ancestral).
FAQ
Comments (12)
🤯🤯
I agree completely with the description, this is indeed much needed now. I appreciate a lot that we finally got this so soon. Thanks a bunch. ♥
I was debating whether or not to share it, as I've only made a few janky LoRA's for my own use with LTX 2.3 (I'm very new to LoRA training and the proper ways of doing things). But it just turned out so good that I absolutely had to share it since I was finding very few flaws in my testing.
Krea 2 really takes to training incredibly well. It can even do viewing angles that I did not include in the training dataset. Also it only took me about 8 hours to train this to 3,000 steps with an RTX 3060 (using Ostris AI Toolkit,). I trained it all the way up to 5500, but 3,000 was definitely the sweet spot.
Glad you find it useful 😀
It even fixes the model's tendency of adding unwanted pubic hair. Thanks again. ✌️
@Kerstal Haha, yeah, Krea 2 really does love peppering the nether regions. It's pretty easy to prompt that out most of the time with something like "hairless crotch", but since none of my training images had hair even running it at a low strength might deal with that problem effectively. Unintended secondary benefits FTW 👍
I'll almost certainly be doing another training run for a V2 and I will be experimenting with a lower rank to avoid unnecessary details being fit onto the LoRA (by training at a much lower rank). So this unintended benefit may go away if I am successful, but the overall quality and detail of the generated images should theoretically go up (closer to base model quality).
First, a good idea and implementation. This LoRA does have a side effect though: it makes the person's skin appear much smoother, practically 3D-like, unless you use yet another LoRA to counter that. I guess it is hard to avoid when using a synthetic dataset, or at least the training dataset should've included samples with varying skin imperfections.
I appreciate the feedback. The more I test and compare with other LoRA's (or no LoRA's) the more I realize it does seem to overly smooth skin out and there is a very noticeable drop in detail/realism. Since most of the dataset images came from SD 1.5 generations, I guess it makes sense. I trained with rank 8 (and tightly cropped) to try and avoid it learning other unnecessary details on the images, but even at that rank it appears it fits more than it should with such a basic concept like this.
Maybe I'll try a quick rank 4 train or perhaps even a rank 1 to really force the LoRA to only learn the absolute basics (no skin details or any of that, just the simple geometrical curves that need to be applied).
I was waiting for this lora since the moment Krea 2 came out. 😊
I recommend 0.77 weight. This lora works pretty well, however I am currently curating a dataset for a similar lora, pushing full realism. If I think it is in any way better/has it's own use case I will upload it eventually.
Do upload it! The more diversity the better. This lora here is decent, but it goes after the polished/glamour look, smoothing the skin details. If you have the "full realism" version, that'd be a different use case entirely, well worth uploading.
I'm working on a new version that hopefully will be as close to the Krea 2 base model quality in terms of realism/details. I didn't realize how much the current version of my LoRA was crushing the details until I compared against the base model (or even a LoRA like Mystic), and the difference is night and day 🙁
I'll be making two big changes with V2, first I'm going to polish up all of the images in my dataset to enhance the skin details/lighting/shadows/etc as much as possible for each one. Then I am going to train at rank 32, as using lower ranks does seem to make things more 'plastic'. I tried doing a rank 4 lora with all the same images and it just made things worse. So I'm hoping that enhancing all the images + training at rank 32 will be a marked improvement. Fingers crossed! But if you do make a innie lora that is more realistic, definitely share it! Until m99 releases one for Krea 2 we have to figure this out on our own.
Also I've found that using dpmpp_sde/beta can help add some skin details, especially at higher resolutions (2048x2048), but it just takes a while to sample.