CivArchive
    Crystal Cauldron - v1.0
    NSFW
    Preview 1248306
    Preview 1248353
    Preview 1248355
    Preview 1248365
    Preview 1248379
    Preview 1248387
    Preview 1248397
    Preview 1248412
    Preview 1248414
    Preview 1248415
    Preview 1248419
    Preview 1248423
    Preview 1248427
    Preview 1248432
    Preview 1248436
    Preview 1248440
    Preview 1248444
    Preview 1248445
    Preview 1248449
    Preview 1248458

    I'm uploading this at the request of user 'oofy' who liked some of the images I shared from it.

    This is just an experimental merge, not really a highly polished mix, so if you want consistency try something else.

    That being said, I think it can produce some very detailed and dramatic looking images with high contrast and vivid colors. It seems to default to a hyper-realistic 3d rendered type of output, but it can be pushed into other styles.

    However, this mix is not without its problems. Known issues include: Producing multi-pane images without asking it to, adding unwanted borders, making horrific 'double faces', and of course... hands. But If you're willing to put up with its quirks, and like the example images, give it a try.

    No refunds.

    NOTE: For some reason the pruned version has noticably reduced color and contrast, and lacks some of the details. I recommend the full version, but some may prefer the pruned.

    All example images were made without any LoRa, inpainting, controlnet, face restore, further upscaling, etc., and mainly just a stack of negative embeddings which I will link to at the bottom.

    I do highly recommend trying it with various LoRa as they can improve the consistency a lot.

    Most of my testing was done with the following settings, so try these first:

    Sampler: UniPC

    Steps: 20

    CFG: 7

    Size: 512x768

    Hi-res fix: 2x Latent

    Denoising: 0.7

    Clip skip: 1

    VAE: vae-ft-mse-840000-ema

    The exact merge recipe is unfortunately forgotten, but I can say with some confidence that it at least involved the following models:

    Orusium by rMadart, RunDiffusion FX Photo and 2.5D by RunDiffusion, ICBINP by residentchiefnz, Juggernaut by KandooAI, Dreamshaper by Lykon, Project Cryengine by hotfantasyai, aZovyaPhotoreal by aZovya, zPhyrNoise by Z_phyr, and (possibly) TerrorYamer2 by Yamer

    I also mixed in some % of one of my older merge experiments, which itself was based on previous versions of Dreamshaper and Project Cryengine, Realistic Vision, Eidomode by Jim Poison, and (possibly) HorrorS4w3d0ffBlend and duchaitendarkside, plus several others that I sadly can't remember.

    Thanks to all of them and any I may have forgotten.

    This model also features a unique blend of essential oil fragrance and the latest in water vapor diffusion technology. :)

    Negative embeddings:

    easynegative

    bad_prompt_2

    NG_DeepNegative_V1_75T

    bad-image-v2-39000 and bad_quality

    Description

    Checkpoint
    SD 1.5

    Details

    Downloads
    982
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    6/22/2023
    Updated
    2/26/2026
    Deleted
    9/22/2025

    Files

    crystalCauldron_v10.safetensors

    Mirrors

    Huggingface (1 mirrors)

    crystalCauldron_v10.safetensors

    Mirrors

    Available On (1 platform)