CivArchive
    Preview 1
    Preview 2
    Preview 3
    Preview 4
    Preview 5
    Preview 6
    Preview 7
    Preview 8
    Preview 9
    Preview 10

    Last updated - 15/07/2025

    Model description:

    This is an attempt at an all-in-one lora for generating the monster girls of MGE, with a focus on customisability, which allows you to select the features you want individually. Refer to these articles for more information. Not everything works perfectly, so good results may require a bit of experimentation at times.

    Background info:

    "MGE" Refers to the monster girl encyclopedia book series by Kenkou Cross. For more information, see the MGE wiki here.

    "Non-entertainment" usage:

    I've had my first question regarding legalities and commercial use, so: I am in no way affiliated with the creator, nor anyone involved in anything regarding MGE. This is a solo project, created solely for fun, which I've uploaded here because I assume other people will also enjoy using it. I haven't, nor do I intend to make any money with this beyond the buzz on this site, and I don't condone anyone else doing so*. I don't feel that strongly about it, and I appreciate being asked, just be aware that if you try I'm not getting involved in any disputes.

    *So it turns out you can now exchange Buzz for real money... This section was written in the context of external projects like pay-walled Patreon content, or using generated images in larger "commercial projects", both due to a lack of legal knowledge, and feeling like it oversteps a boundary with my personal ethics. I don't intend to change my stance on this, as I think the negative online perception around generative AI stems from the immediate rush to monetise it, and I think most "ethical concerns" would have taken years to be brought up, if ever, if things like ChatGPT remained as niche research projects.

    Actual artists:

    If any artists dislike the idea of their stuff being used for this (and you've drawn things from MGE), let me know and I'll avoid your work. I understand that AI is a very divisive topic, and I didn't make this with the intention of ??????? people off. As I don't intend to make money using my model, thid is more-so a thing for artstyles, because obviously impersonators will become harder to distinguish the better AI gets.

    Reuploads:

    I've decided to prevent unrequested reuploads of this (and my other stuff) on other sites as I've found the claiming process to be quite tedious, and there's little to no reason to allow it. I'll take direct requests for uploads, as if there's a reason to put it somewhere (say, for instance, a new startup has really good servers/features), then there’s a valid reason for doing so. I'm just not comfortable with stuff I've spent a significant amount of time making being spread to places I'm not aware of, nor people that did nothing to help benefitting from it. If you want to spread it elsewhere through links or something, go ahead, just don't reupload it to get yourself some free credits.

    Description

    (Even bigger file size this time, sorry, I used 160 dim / 32 alpha)

    This version is mainly to fix the things V5.1 struggles with, and should (hopefully) give more consistent results. Also, I randomly decided to tag the raiju's legs as "calfhigh fur" instead, if you're after that specific leg style.

    The style also works a lot better now, if you use "mge artstyle", and obviously higher weight increases the strength. From my limited testing, the style tends to give less detail the stronger it gets (see the newest manticore/dark slime images), but I'd assume using better prompts, different models or other loras should all be able to mitigate that.

    Things I'm aware of:

    - General inconsistencies - Fixed, see model description for tag order fixes

    - Furless bunyips

    - Winged arms not working

    - "Full body" significantly reducing detail

    - "MGE style" decreasing fidelity

    - Slime girls always being opaque - (Sort of) fixed by tag order changes

    - Gremlins having green skin

    - Manticore tails being bad (this is likely to stay here for a while...)

    - Characters having inconsistent features/clothing across multi-scene images (doujins, refrence sheets, etc)

    - Smeared eyes/mouth (try "smeared eyes/mouth" in negatives, potential fix for no/bad pupils.)

    - Raiju hair color inconsistency (which the AI kindly reminded me of by itself.) - Fixed, use tags: gradient hair, glowing hair, "x" hair to "y" hair (x and y being colors, scalp to tip)

    - Bad barometz fruit (girl herself tends to be good)

    - A few people have requested smaller file sizes, so I'll look into that too. Bear in mind, I'm a free user that can't train models locally (afaik, that is. If anyone knows of an AMD-compatble kohya alternative, lmk), so it'll likely be a while before I get anywhere with this.

    FAQ