CivArchive
    DT_Realmax photography model - V1.0
    Preview 1
    Preview 2
    Preview 3
    Preview 4
    Preview 5
    Preview 6
    Preview 7
    Preview 8
    Preview 9
    Preview 10
    Preview 11
    Preview 12

    Description

    Let's expand and answer the raw picture questions: ” data-href=” https://liblibai-online.liblib.cloud/img/e69f8285652d4dfda393e4473f637549/764e62a09ff2f6efa52361074418e990bb3e5a3fd66cebafad022ace082a5663.png?x-oss-process=im age/resize, w_1290, m_lfit/format, webp” style=

    Retraining+fusion of realistic style models because someone wanted to have fun, so they did it... The parameters are set according to Dgirl. Beginners are

    not recommended to add a bunch of embeddings, which will affect the texture. When drawing, don't throw a bunch of lora inside; on the contrary, the effect will be much better. When drawing, don't worry about adding a bunch of things; try every element first to see if the model can be made. (Old player, I didn't say that

    ) Let me know if it's a fusion.

    Negative please use this: (worst quality: 1.6), (low quality: 1.6), (normal quality: 1.6), lowres, watermark, nsfw,



    1. About repainting : Similar to portrait, close_up, this type of close-up description requires 0.1-0.3 high-definition restoration, or even high-definition restoration. The number of sampling steps is arbitrary (anything that is ridiculously low or around 100 steps), and the number of high-definition restoration steps can be 8-25. People have become fake Number of steps Just A little less , it would be nice to redraw it a little lower. The picture is too blurry or blurry Instead, turn it up a bit.

    2. Using ADetailer : You can't just leave AdTailier there; you also need to set the redraw amplitude, For example, your face is fake and very oily

    , AD redrawing needs to be adjusted to about 0.3,

    ” data-href=” https://liblibai-online.liblib.cloud/img/e69f8285652d4dfda393e4473f637549/5658dddfa9664aec6d21289f54355b0928960b2b1a622c95dc2b44d0d33fb501.png?x-oss-process=im age/resize, w_1290, m_lfit/format, webp” style=

    Number of steps: 20,

    Sampler: DPM++ 2M Karras, (others are OK, depending on personal preference)

    Cue Correlation CFG scale: 7,

    Size: 448x800,

    VAE: kl-f8-anime2.ckpt, or not used

    HD redraw amplitude Denoising strength: 0.48,

    Number of iterative steps for a high score: 15


    (The view is higher, so as long as it's not too fake, you can raise it appropriately)

    Clip skip: 2,

    FAQ

    Details

    Downloads
    178
    Platform
    ShakkerAI
    Platform Status
    Available
    Created
    3/31/2024
    Updated
    5/29/2024
    Deleted
    -

    Files