In my quest to find that "consistent" but "flexible" workflow for storytelling, game creation, and the like, I started playing with some new ideas and concepts. I came to the realization that if I paired a "Consistent" model with sliders/LoRas that could be used to "step through" some transformation and fun sequences like that - it actually worked well. So I started playing with this. (Backstory and ingredients are listed at the end).
What Is This?
XenoMORPH is a checkpoint that is both flexible and efficient at both portraits and scenery PLUS, it has been optimized for use with other LoRas and Sliders so that it holds the style and general aesthetic as best it can no matter what you throw at it. Many of the popular sliders and morphing type LoRas have been merged in here at VERY low levels to start out - so they don't really have much effect on the model, but then when you actually decide to "use" that LoRa, the checkpoint has already been "primed" so to speak. This means that changes to styles and outfits as you "morph" your subject into other things change a lot less since it's already got the idea from that initial "low level" merge as to where to go. In many cases that worked well - better than expected really. And in other cases, it doesn't seem to make a huge difference, but it's primed nonetheless.
That means that:
It should work well with most (decent) LoRas, Sliders, Embeds, etc.
It pushes toward a realistic look, but style and aesthetic prompts and LoRas might change the "shape" of things (at low levels anyway) without affecting the actual style. It reduces the tendency to, for example, go from a dark dungeon to a Barbie dreamhouse as you use your favorite transformation/slider type LoRas at different weights. It's not perfect, but does a REALLY good job, IMO.
Handles simple prompts REALLY well, and does a pretty good job with (most) more complicated prompts. I did find a few avenues where things get a bit weird - but that's true with just about every checkpoint I've ever tried.
Need Help? Want Tips, Tricks, Pix of Things We Can't Show on Civitai?
Join the XenoEngine Discord to discuss tips and tricks, complimentary Lora's and Embeds, and to post (and see) images not allowed here on Civitai.
Basic Aesthetics
You can get to other styles with this, but it really want to stay fairly realistic looking as much as it can. To change styles you'll want higher weights in the calls than your normal checkpoint. The plus side here is that you can often bring in some of the "shapes" and "colors" and other aspects of a style long before it "breaks" the reality look. So, for example, if you use my LucasArts Embedding here - you're get some of the fun shapes, hairstyles, clothing, background types and so on before it starts to actually get to it looking like "art/cg illustration".
Larger Default Breast Size: One of the bigger tricks with offset breast sizes (prompted or with help of many sliders) is that the further you travel from the baseline, the more the style of the training data kicks in. By making default breasts larger - closer to the "middle" of that scale - the very small and very large sizes have a smaller offset difference and therefore don't fight as hard to change things. It's not perfect, but really helps a lot.
NOTE: Want smaller breasts? - check the "BreastsOnly" XenoSlide from the section below. At negative numbers, you can dial them down and have very little effect on clothing or scenery.
The Companion XenoSLIDEs
At this point, there's just one set, but I hope to add more for various other things as time goes by.
XenoSLIDE - Bimbos & Boobs: 4 different sliders starting from a basic "tiny to larger" breast sizer all the way up to a full "Amazon Bimbo" type. See that page for more info.
Backstory
If you know me and/or the various models I've released over my tenure here on Civitai, you probably already know that I tend to enjoy transformation, corruption, and mind control themes. I'm also looking to have ways to "tell stories" in games, comics, or however else one might want to do that - but it's always hard to keep consistency from generation to generation.
One of the biggest challenges of body transformations (such as bimbofication and the like) is that as you push toward your target, it also ends up changing the environment and the whole aesthetic of the image. If you've ever played with most Bimbo type Slider LoRas, you'll know that once you get up to the top, it can be hard to do things like avoiding pink walls and clothing, getting everything so it doesn't scream "Barbie World!"
And so, this week's mission was to attempt to create something where you could have the widest range of flexibility, lots of options, and the least amount of impact on the overall image generation style across multiple generations. I started playing with a Slider Making Colab along with various model merges and seeing what I could do. Would it end up being a checkpoint? A Slider LoRa? Or... Both? (With multiple slider variations to suit various tastes and body types.)
The answer, was... Both.
Through LOTS of experimentation, I got to playing around with merging all kind of different things, but with a "slider" type concept, that's not going to work in a checkpoint. So I set out to make a checkpoint that simply LOVES Loras of all types and then sliders (one of my own creation plus merging in other sliders at specific weights to get specific results).
The Checkpoint
A realistic base model but with a little extra "bustiness" added to the default character. (I'll get into the "why" of this in the "Reasoning" section, below.) I started by finding Plastic Reality by @Fizzlebomb3r. It is a bit overkill for my liking BUT... it looked like (aside from clothes tending to always be very low cut) it had a much easier time of putting people into the outfits you want than with most "bimbo" type things which just did a limited set of "slutwear" items. I then discovered Daring Hydra by @pyn - and I loved that the merge combined some of my favorite "portrait" models along with a bunch of "scenery" type things, too. And the overall aesthetic hit right where I like it.
So... Plastic Reality got merged into Daring Hydra at a low weight to boost up the starting breast size. By doing this, I hoped that since the starting breast size was larger, "offset" sliders would end up starting with a clothing set already keyed toward larger breasts and it would cut down on the "Business Suit becomes a Latex Catsuit" effect.
Then I merged in some of my XenoGRAPH and XenoGASM checkpoints and ended up with this beauty.
Description
Added some Hyperfusion and some elements from XenoGRAPH. Slightly smaller starting breast sizes due to the top end of sizes being much larger. A little extra skin detail and hopefully some improvements to hit/miss ratios with hands and eyes.
FAQ
Comments (11)
I've been getting the following issue when using v2 in Easy Diffusion:
"Error: Could not load the stable-diffusion model! Reason: 'Attention' object has no attribute 'to_to_K' Task ended after 10 seconds."
v1 works perfectly fine though. Did you change how the model is merged or packaged or something? This is the first time I've ever encountered this kind of issue with any model.
Hmmm. No. No changes...
Give me a few minutes to boot up an instance - I'll download from here instead of using my local version and see if it got corrupted during upload or something...
Yeah upload corruption could be a cause, but I also wonder if it could have to do with merging with the new hyperfusion model. Idk, but those are the first two things that come to mind
Okay... worked fine for me on A111 downloading from the site here - so it's not an upload issue. (Plus I would imagine I would have heard something before now if it was).
A google search led me here (with Easy Diffusion being the first result).
https://github.com/easydiffusion/easydiffusion/issues/1618
The link in the response there takes us to something about diffuser bugs - supposedly fixed. Now--- I've not used Easy Diffusion... so I'm not sure the story here. The bug is "closed" on that second page - though it also has a fix for people to paste in too - so I'm not sure what's up. A non-compatible model in my mix that screwed up Easy Diffusion? I really don't know and am not really up on all this technical stuff...
Will try to do some research. Sorry you're having trouble.
Ohhh. Here's an idea...
My XenoEngine Artstyle V5 (https://civitai.com/models/46049/xenoengine-artstyle-mega-model) has the same "new stuff" module as this one. The rest of the mix is quite different and it's tuned a lot differently - but in essence... if V5 also doesn't work for you... then we might be able to determine if it's just a general error, or something in that new mix and how it got merged. (You don't need to bother with this if you don't want... but it might help me get a handle on it).
Again... sorry you're having headaches.
@iamxenos Thanks for delving into this. I know this is a relatively niche issue since Easy Diffusion is not the default compared to A111, I just prefer the interface and find it a lot more accessible/easier to install than A111 or ComfyUI, while still allowing for plenty of customization and support for multiple models. Switching to the V3 engine in Easy Diffusion's settings helped immensely too. Plus now that it supports using multiple LORAs, Controlnet, and has it's own model-merging feature, it achieves virtually anything I need in terms of image generation. Here's a link to the main download for V3 for you and anyone else who want's to try Easy Diffusion for further testing:
https://easydiffusion.github.io/
I have used Xenoengine V5 but while I think it's functional for my prompts and flexible with the most of LORAs I use, Xenomorph is a lot better at facial accuracy (especially when using TIs for a specific face) and being more true to my prompts in general. Also Xenoengine leans a bit more towards a western cartoon-style in combination with my prompts/LORA setup, while Xenomorph excels in hyper-realism, which I prefer.
Again, no issues with V1, but somehow something got borked with V2. Here's hoping future versions of this model somehow inadvertently fix this, because I really enjoyed V1 of Xenomorph and would like to see how else it can be improved/modified in the future.
@Kodama Okay... my main thing with XenoEngine V5 was just to see if it was something in that new merge pack or not. And since that worked for you - it's almost like it's a corrupted file problem or something. Or else you'd get the same issue with the XE-V5.
@iamxenos Yeah that's why I'm absolutely stumped. I even tried re-downloading Xenomorph V2 to see if that would somehow fix it (to rule out any download corruption issues), but no still the same issue
This isn't photo-realistic bud.
Trying to see in the description where I said it was photo realistic. Semi realistic or 2.5D is probably as close as I'd go - but I could have made a mistake somewhere in there... I'm just not seeing it. (Chances are I DID make a mistake, really... but... help me find the spot where I'm calling it that, please).
The tags. Which is what I was searching for lol :P
@Twinmind00 DOH! You're right, I'm an idiot. I have no idea why that tag got added. Probably a few two many beers on launch day! lol Thanks for spotting and pointing it out.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.

