CivArchive
    Preview 116918854
    Preview 116920531
    Preview 116919441
    Preview 116918858
    Preview 116918855
    Preview 116918861
    Preview 116918848
    Preview 116918860
    Preview 116918856
    Preview 116918849
    Preview 116918857
    Preview 116918852
    Preview 116918851
    Preview 116918862
    Preview 116918850
    Preview 116918853
    Preview 116918859
    Preview 116919821
    Preview 116919916
    Preview 116925327

    A reboot of Sibyllic into realism vs fantasy. This series will focus on erotic art and fantasy themes. It is a break from v1-v5.

    • Fine Art

    • Cinematic Fantasy

    • Erotic Art

    • Satirical Concepts

    • Unusual Surrealism

    I look forward to seeing the creations you can come up with.

    If you like what I do, I now have a link for tips. You can also use the link below to send me a private message or model requests:

    Tip Me

    Description

    BF16 mixed precision & fp32 clip

    My settings:

    DPM++ 3M SDE SGMUniform

    40-50 steps

    5-7 CFG

    FAQ

    Comments (6)

    tikisquidJan 10, 2026· 1 reaction
    CivitAI

    A workflow would be helpful, trying to use this in comfy, using your recommendations, produces a lot of body horror and strangeness. I asked for a girl with a wizard hat, it gave me an old man with a beard...

    AKDesignsJan 22, 2026· 3 reactions
    CivitAI

    SDXL model with 9Gb is too big, almost size of Flux model :(

    Can you make smaller model FP8 or Q GGUF version???

    IntosicaJan 22, 2026

    From what I remember, I think you might still able to load on 8gb VRAM

    AKDesignsJan 23, 2026· 1 reaction

    @Intosica Yes I can Load it & run it BUT just like Large Flux model it takes much too long to generate 1 image 40-50min @512px*512px with 4-step lora :((

    Smaller models run faster on my ancient hardware!

    + it takes less storage space on limited size HDD !!

    IntosicaJan 24, 2026

    @AKDesigns That is definitely insane I'd say.
    The new Z-Image quantized has some very interesting performance on my shitty 8gb GPU.
    5 minutes on initiating, and then like less than 90 seconds for a while and then if you dont touch it for it a little bit, it takes 5 minute initiations.
    I know it's not SDXL, but certainly a better alternative if there is any out there in 2026 for us small GPU guys.
    Was super impressed what it text encoder could do on my personal PC.

    Edit: I lied. It's actually even quicker on my 3060ti
    even loaded the model super fast.
    Same goes before, but 70 second initiations and 10-15 second gens.
    No high-res fix though.

    hkg364Feb 22, 2026
    CivitAI

    Very good checkpoint. Adetailer is usually not needed, and has to be put at a lower inpaint denoise to not make it look too smooth.

    Checkpoint
    SDXL 1.0

    Details

    Downloads
    744
    Platform
    CivitAI
    Platform Status
    Available
    Created
    1/10/2026
    Updated
    5/12/2026
    Deleted
    -

    Files

    sibyllicXL_circeV6.safetensors

    Mirrors