CivArchive
    XenoXTC (Art/Toon Portraits and Fetishes for CG Game Art) - vCMX
    NSFW
    Preview 5585473
    Preview 5585520
    Preview 5585519
    Preview 5585517
    Preview 5585522
    Preview 5585560
    Preview 5585559
    Preview 5585561
    Preview 5585564
    Preview 5585562
    Preview 5585563
    Preview 5585566
    Preview 5585565
    Preview 5585567

    With my XenoGASM model reaching the point where it's able to generate just about as real as it can get, it's time to start turning things back the other way - heading back toward a Western Digital CG sort of art style.

    For years, I've been looking for a way for me to either reboot or continue the stories I wrote and posted at MCStories.com almost 2 decades ago as a game or piece of interactive fiction/visual novel. I wanted a way to be able to define my characters and (relatively easily) put them into different outfits and locations without having to spend 4 hours prepping a scene in DAZ Studio or similar. Other character tools like iClone look fun on the surface, but making new outfits and scenes is even more tricky.

    On top of that - I enjoy allowing players to customize and mod games they enjoy so that they can bring some of themselves info the experience - so I wanted something where someone else could just create a character look and be able to easily match the styles of the game and insert them in as if they were a part of the game all along.

    XenoXTC represents the milestone where I think I've got a great game character and basic environment model and it is time to start the other side and begin putting together my years of code snippets to actually start making the game.

    XenoXTC

    This model is still an evolving process (see intro, above). It started out as this latest version of XenoGASM and then I mixed in a merge I've been working on for quite a while - a sort of realistic look and feel but with very smooth, maybe even plastic, skin. (Eventually I'd like to get the output default to be this sort of look, but with shapes and forms - both in people and scenery - that hearken back to the old LusasArts Titles.

    See the initial pics posts for various comparison and prompting ideas and suggestions.

    vCMX

    This is another style variation with the same FreeForm base as vTOON. Less Disney/Pixar big eyes and shapes, more Comics and flat shaded styling. It does not replace other versions - just another styling option.

    VToon

    This is the version that will continue to be developed. At the initial release time (Nov. 19, 2023) the model has its base style and generates a very diverse range of default "1girl" variations and also takes known entities like Celebrities (or Chimera Combinations - e.g. [Molly Ringwald|Christie Brinkley]) and does great portraiture with simple or detailed backgrounds.

    It's NOT great at doing "action" type scenes yet - where both the characters and scenes have lots of detail - the scenes tend to suffer and get a little warpy. I'm working on some merges to help with that. The main idea is to create sprite character and bust graphics (for Visual Novels or RPG type games on RPGMaker and similar) so I'll typically be extracting characters and overlaying them as sprites over different backgrounds anyway - but I still want to improve this so that cool self-contained cut scenes will work well, too.

    The one thing I haven't, yet, been able to really accomplish is getting the exact skin look I want - so you'll notice I often add smooth skin and glossy skin to the prompt. These give it that extra polish. Sometimes it works well without and these can be overkill, which is why I didn't force those concepts into it. Plus those making harder to get wet skin looking nice.

    Calling for clothing items along with a style (e.g. Steampunk Style Outfit, pleated skirt, long sleeve buttoned blouse, penny loafers) gives a nice on-theme variety of outfit looks for which we'll eventually get some workflow pipelines to help keep things similar across generations for the same character in the same outfit. Not sure we can get it as perfect as we'd like for normal game production... but we'll see how close we can get. Stay tuned.

    V1 - FreeForm

    This is a more free form version before the styles and additional concepts have been merged into it. It is mainly here for others who want to build "bridge" merges (combined with a Style Slider Lora to easily "morph" between styles so their characters might explore virtual worlds or alternate universes within the game which consist of entirely different looks and feels. (I have a few fun game mechanic ideas I'd like to play with here, too. We'll get into those in a week or so).

    There is a Full (Just under 4gig) version and a normal 2gig pruned version of FreeForm available.

    General Usage Tips

    The TOON version fights hard to keep its style or at least hold onto the general character aesthetic. This is done so that we can do things like describe a "Steampunk" outift or even a "Skyrim Style" outfit without the model also changing the whole character and feel. In many cases it works, but some strong styles and calling specific artists may not work as well - depending upon how well this style and your prompted style play together.

    Simple is better - at least to start. Complex scenes are always tricky, but on a model like this - they can be even trickier. So if you're trying to create a complex scene and things are going bad with your long prompt - the way forward can often be to remove things more than adding more tokens to push back.

    I'll be updating here and posting articles and tips on the discord, so... stay tuned.


    I'll update this over time (and likely start a Dev Diary Article or something). If you have ideas, suggestions, or if you would like some help on how to prompt for a specific look, pose, or action - feel free to drop by the Discord and ask. (Link below). I've started a collection of openposes for portraits and busts that you can find in the resources section of the discord, too.


    I'm using Civitai's trainer - so Buzz Point tips are helpful - even just a few add up.

    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.



    Description

    This is basically the same as earlier versions, but with a more "comics" and less "cartoon" style than vTOON. It's still going to be best for character portraits/busts than complex scenes for storytelling. It also does a decent job of backgrounds, but combining a character with a background in the same generation can have some style shift happening in the backgrounds. (i.e. if you're making a game, I'd recommend generating a background image separately from your PNG/Transparent Sprite character overlays.)

    FAQ

    Comments (10)

    maybellesunJan 17, 2024· 1 reaction
    CivitAI

    Yay, the earliest christmas this year! I am so looking forward to play with this gift!

    maybellesunJan 17, 2024
    CivitAI

    Dear Xenos, it seems to me, this is an LCM-Model, and i just love it.

    But i am having the same problem with this one as with my own... i cannot think of a prompt making it useable when NOT using LCM. Any Idea how to transfer the LCM-Result style into "slow-mode"?

    iamxenos
    Author
    Jan 17, 2024

    I have NOT created an LCM version of this yet.

    You CAN use it with LCM if you get this LoRa - https://civitai.com/models/195519?modelVersionId=225222
    That is what I did for most of the previews.

    Otherwise, without that LoRa - you can use normal CFG and Sampler Settings (e.g. Euler A, 20 steps, CFG scale 7) too.

    I should have the LCM premade version ready in a few days.

    maybellesunJan 17, 2024

    @iamxenos why everyone misunderstands my problem....

    with LCM sampler and LoRa (baked in or not) and the corresponding settings-> i can prompt whatever i want, it does at least not look bad. (even empty prompts give a nice picture, i would call this the checkpoints "LCM-Default-Style")

    with every other sampler and the corresponding settings -> i need a "baseprompt" to push the checkpoint in an at least ... not complete ugly... style.

    and i have NO idea what words to use to achieve something that at least look a fraction as good as the "LCM-Default-Style"!

    In other words... its not the Rendersetting, its the words / Prompt i struggle with.

    iamxenos
    Author
    Jan 17, 2024

    @maybellesun How many steps are you using with you have the LCM sampler and LoRa NOT in use?

    iamxenos
    Author
    Jan 17, 2024

    Oh - and your CFG Scale too - that should be up above 4... I like to use 7 normally.

    maybellesunJan 17, 2024

    @iamxenos okay.. one more try.

    Create a picture that captures the style of the Checkpoint.... WITHOUT LCM-Sampler or the LCM-LoRa.

    Thats where i struggle.

    iamxenos
    Author
    Jan 17, 2024

    @maybellesun Ahhh... I see - the style changes... gotcha. I have a meeting in a few minutes, but I'm going to start a batch of tests - empty prompt, 1girl, and woman as the only words in the prompt - and then we can look at them and see.

    Stay tuned... might be a few hours before I am out of the meeting and have time to look my results over.

    maybellesunJan 17, 2024

    @iamxenos yay, you got it.... finally ;) Since i have a potato-PC ,i created a checkpoint myself only checking it with LCM until i liked it.... but i did not manage to use it in "slow-mo" (funny thing, my checkpoint is heavily inspired by you - even though nothing of your resources were baked in).

    I can hardly wait to see the solution you come up with!

    well on the other hand... i will go to bed soon, so take your time ;)

    iamxenos
    Author
    Jan 18, 2024

    @maybellesun Sorry - didn't get a chance to look at these until now...

    Ran these tests and I don't really see any issues. https://civitai.com/posts/1219108

    The negative embeds I was using (Badhands and BadAnatomy) seem to help with colors a bit, but even without them - the style seems to be there. I'm not really sure what you're seeing on your end.

    Checkpoint
    SD 1.5

    Details

    Downloads
    1,269
    Platform
    CivitAI
    Platform Status
    Available
    Created
    1/17/2024
    Updated
    5/24/2026
    Deleted
    -

    Available On (1 platform)

    Same model published on other platforms. May have additional downloads or version variants.