CivArchive
    ArtCore - v1.0
    NSFW
    Preview 18718606
    Preview 18723942
    Preview 18849686
    Preview 18724889
    Preview 18708079
    Preview 18710220
    Preview 18708454
    Preview 18710356
    Preview 18593767
    Preview 18713677
    Preview 18710129
    Preview 18724206
    Preview 18708105
    Preview 18722467
    Preview 18713434
    Preview 18713790
    Preview 18726370
    Preview 18456489
    Preview 18720743
    Preview 18713875

    This checkpoint is a blend between brutal realism and art.

    To archive this goal I have used a unpublished version of DemonCore. Its based on version 4.1 MidgardBeast got an additional training on over 2000 images... This version got also my own enhancer merged in. I may will publish this one soon.. Freek22 helped me a lot creating the newer Midgard checkpoints.

    The other part of the model is Kika_real V4 https://tensor.art/models/745584530192150460 an awesome model that is a bit more art focused and also worth checking out. This model was created by speach1sdef178 and Triple_Headed_Monkey

    ArtCore is intendet to create realistic images with almost HDR like colors. Its very creative for a realism checkpoint. Feel free to try different styles though. It can also produce NSFW things but this model is not intented to create porn.

    Settings:

    I will add a official workflow for it soon. Most of the images are created with Forge. I recommend using around 15-25 steps (tested with SDE) and around 2-6cfg. It workd below that but you will get more details using this setting.

    Nearly all samplers are working with it. I don't recommend using Heun (normal), 2M SDE (normal) and 2M SDE Heun (normal).

    I will include the images I created with Comfy in the training data. Here you can get some basic workflows. Just load the images into Comfy.

    Description

    I recommend using around 15-25 steps (tested with SDE) and around 2-6cfg. It workd below that but you will get more details using this setting.

    Nearly all samplers are working with it. I don't recommend using Heun (normal), 2M SDE (normal) and 2M SDE Heun (normal).

    I will include the images I created with Comfy in the training data. Here you can get some basic workflows. Just load the images into Comfy.