CivArchive
    Preview 45139366
    Preview 45139368
    Preview 45139381
    Preview 45139378
    Preview 45139379
    Preview 45139384
    Preview 45139370
    Preview 45139376
    Preview 45139399
    Preview 45139388
    Preview 45273721
    Preview 45289414
    Preview 45294188
    Preview 45300801
    Preview 45306246

    Regards for all your love and support ❤

    avoiding negatives can improve result on SDXL

    Description

    be stunned

    FAQ

    Comments (16)

    highfalutinDec 13, 2024· 1 reaction
    CivitAI

    Very noticeable improvements across the board in VXI, especially in face shape, variety, and skin detail! Much better softcore NSFW as well. I am indeed thoroughly stunned.

    (One small bit of critique about this line of models though: specifying bust size has much more of an impact on overall face shape and body type than I think most would expect, and those effects are in most cases very stubborn)

    creedukDec 13, 2024· 7 reactions
    CivitAI

    You just convinced me to pause the SD3.5/Flux gens and go back for some SDXL fun

    pra111122223333951Dec 14, 2024· 4 reactions
    CivitAI

    thanks for keeping v8 still in the generator

    epinikion
    Author
    Dec 14, 2024

    I know you guys 😅

    evart304Dec 14, 2024· 6 reactions
    CivitAI

    Amazing work, incredible results. Your model is one of the best. Thank you for your hard work

    3567304Dec 14, 2024· 1 reaction
    CivitAI

    waiting for 4 step lightning version

    DelightofVisionDec 18, 2024

    I waiting too

    6vidit9Dec 16, 2024· 4 reactions
    CivitAI

    This VERSION is INCREDIBLE🔥❤️

    epinikion
    Author
    Dec 16, 2024· 1 reaction

    Thank you mate! As always, have fun creating ❤️

    Off2OnDec 16, 2024· 4 reactions
    CivitAI

    excellent fidelity, but impossible to generate big naturals

    alexoterinDec 17, 2024· 1 reaction
    CivitAI


    I wanted to do a human lora training using epicrealismXL_vxiAbeast, but the model does not load into kohya and gives an error. Is your model built with errors?

    Corleone111Dec 17, 2024· 1 reaction

    TBH, use the base SDXL model. I've trained many, many Loras, also with different checkpoints, but the best and most flexible results are still achieved with the base SDXL model.

    Corleone111Dec 17, 2024· 5 reactions
    CivitAI

    I don't know about the other versions but the latest (aBEAST) doesn't work well with Loras.

    epinikion
    Author
    Dec 18, 2024

    which Loras?

    6vidit9Dec 18, 2024· 1 reaction

    It is working dude, I have used a few Loras and it is working perfectly well for me at least.

    Corleone111Dec 18, 2024

    @epinikion Some of my own that I created with the SDXL base model, that work flawlessly on any other checkpoint. But It might be some general settings that I have to tweak on my side for the generation in general. I'm currently using around 40-50 steps with 2M SDE or Heun. Is there something else I have to pay attention to (using Forge)?

    Edit: There also seems to be a lot of noise and artifacts that aren't present in other checkpoints. The same problem happened in NatVis 2.0 but the creator said that something went wrong with the text encoder during the fine tune.

    Checkpoint
    SDXL 1.0

    Details

    Downloads
    14,393
    Platform
    CivitAI
    Platform Status
    Available
    Created
    12/12/2024
    Updated
    5/13/2026
    Deleted
    -

    Available On (1 platform)

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