CivArchive
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined
    Preview undefined

    Vox-Mix

    Vox-mix is a checkpoint optimized for creating images influenced by Victorian and Art Nouveau styles.

    * The main sample images were created using positive/negative prompts only.

    No embedding, Lora, A-Detailer or Dynamic Thresholding was used.

    Recommended keyword

    Victorian era cute girl
    Oil Painting by (Francois Boucher, Alphonse Mucha, Claude Monet, Franz Xaver Winterhalter, NORMAN ROCKWELL),

    About Vox-mix2.8m

    Major improvements

    - Improved body balance distortion and reduced the likelihood of generating poorly composed images

    - Fixed issues with NSFW images (extra nipples, incorrect genitalia, etc.)

    - Fixed issues when applying LoRa (face magnification or over-representation)

    A cup of coffee : https://ko-fi.com/voxmix

    Description

    Vox-mix2 Dark Lite

    This version has been partially adjusted from the Vox-Mix2 version to be more specific to NSFW images, and the file size has been reduced.

    A cup of coffee : https://ko-fi.com/voxmix

    FAQ

    Comments (8)

    settima_aiApr 21, 2023
    CivitAI

    the "Novel AI" model where did you find and how is big?

    vox_mix
    Author
    Apr 22, 2023

    I have downloaded a model of Novel AI in 2022, when the AI image was born (around October). The image was about 7GB, but I don't have it at the moment.

    kurodahuga843Apr 21, 2023
    CivitAI

    I found a model that "stabilizes quality even when multiple people are lined up. Thank you very much.

    vox_mix
    Author
    Apr 22, 2023

    Glad it was helpful for your AI images.

    LovelaceAApr 22, 2023
    CivitAI

    May I know how to make a checkpoint model focus on one style, like you did for this one?

    vox_mix
    Author
    Apr 22, 2023

    At first, merging was done using weight merging, but since using Unet, more detailed settings have become possible.

    After creating various models, focusing mainly on the background-based area, create images with background prompts and compare them with x/y pilots.

    It is time-consuming and inconvenient, but you can create a model that reflects your specific taste.

    Hope this helps.

    appleqqqqqqqqqqp5495Apr 30, 2023
    CivitAI

    What is vae? Can't draw this effect?

    vox_mix
    Author
    May 1, 2023

    Oh. There was no form to fill in the VAE information, so I didn't fill it out separately. I'm sorry.

    Basically, the image I create is of hakurei

    It uses kl-f8-anime2.ckpt which is an SD VAE based on waifu-diffusion-v1-4.

    I recommend you try it out.

    Checkpoint
    SD 1.5

    Details

    Downloads
    1,288
    Platform
    CivitAI
    Platform Status
    Available
    Created
    4/21/2023
    Updated
    5/11/2026
    Deleted
    -