CivArchive
    Fallen Art | ILL - Fallen Art V2 | ILL
    NSFW
    Preview 96576402
    Preview 93784887
    Preview 93784877
    Preview 93784886
    Preview 93784860
    Preview 93784861
    Preview 93784859
    Preview 93784862
    Preview 93784863
    Preview 93784872
    Preview 93784867
    Preview 93784864
    Preview 93784868
    Preview 93784869
    Preview 93784871
    Preview 93784865
    Preview 93784866
    Preview 93784884

    I would appreciate feedback

    Experimental: Show me what your images will look like before this version gets/doesn't get its honorary title on my list :D

    V3 Well...

    Finally CivitAI gave me a decent dataset to download for version 3.

    LORA is trained on about 400 images so the style learned from the very first version is transferred even better.

    V2 In that version I tried to convey exactly the style that I get when working with Yiffimix on Illustrious models.

    IMPORTANT for V1! If you want to get +- similar results to mine, then use the yiffymix checkpoint https://civarchive.com/models/3671/yiffymix and ComfyUI

    The principal difference is that this LORA is trained on a specific checkpoint, WAInsfw

    Description

    In that version I tried to convey exactly the style that I get when working with Yiffimix on Illustrious models.

    The fundamental difference is that this lore is trained on a specific checkpoint, WAInsfw

    I would appreciate feedback

    FAQ

    Comments (6)

    xiakezhiguang01Aug 13, 2025
    CivitAI

    pretty good

    Prog420ressAug 13, 2025
    CivitAI

    Notice anything strange about characters' eyes when doing close ups? Did you use Reactor or an eye LoRA when making your dataset?

    Including "cross-eyed" and "bad eye proportions" in the negative prompt took care of it.

    I like the overall style though!

    FallenEnder
    Author
    Aug 13, 2025

    I usually use FaceDetailer and don't have much of a problem with squinting, but I'll add that to the negative prompt for future generations

    whity00Aug 24, 2025

    If you use a face deatailer, at least for me with ADetailer it takes the checkpoint style.

    But you can put the Lora+trigger word into the detailer so the style stays the same + better faces/Eyes

    whity00Aug 24, 2025
    CivitAI

    As of now it works pretty well.

    Im still in the middle of creating something.

    The only thing I see is that it increases the breast size and dimensions of my female characters by at least one notch :D

    But nothing that can't be fixed.

    FallenEnder
    Author
    Aug 24, 2025

    yes, there is such a thing, you can look at the previous version and you will understand why it is so). I have a database of images with different body types, poses, facial expressions, but the site does not want to load them normally for training. So who knows when the third version will come out XD

    LoCon
    Illustrious

    Details

    Downloads
    293
    Platform
    CivitAI
    Platform Status
    Available
    Created
    8/12/2025
    Updated
    8/28/2026
    Deleted
    -
    Trigger Words:
    GakuStyle

    Files

    Gaku_Artstyle_v_2__ILL.safetensors

    Mirrors