santa hat
deerdeer nosedeer glow
Sign In

SilvermoonMix Base model

1.5k
11.3k
168
Updated: Oct 5, 2024
style
Verified:
SafeTensor
Type
Checkpoint Merge
Stats
4,863
Reviews
Published
Nov 19, 2023
Base Model
Other
Hash
AutoV2
3B4D7650D3
License:

这个模型有较好的构图、衣服、手脚逻辑和光影。它有非常干净和中性的画风,这使得它可以搭配各种画风LoRA。查看我已经发布的画风LoRA:https://civitai.com/user/silvermoong/models

This model has great backgrounds, cloth, limbs and lighting. It is designed to have a clean and neutral style to work with style LoRAs. You can find my style LoRAs here: https://civitai.com/user/silvermoong/models

推荐负面/Suggested negative prompts: (worst quality, low quality:1.4)

重量级更新,手部修复,使用方法如下!!Hugh update: Introducing hand fix!! Read more for how to:

使用方法/how to:

Handfix可以直接使用,也可以按照下面的方法配合使用。

Handfix model can be used directly. Or use with the original one together:

  1. 把基础模型切换成handfix版本
    Switch to handfix checkpoint

  2. 在设置中把enable refiner设置为first pass
    Under settings: set enable refiner to first pass

  3. 把Hires.fix和refiner模型设置为正常版本,修改refiner切入时机
    Set Hires.fix and refiner models to original one. Modify refiner switch at based your steps

基于基础28步,高修14步的效果/effect based on 28 steps+14Highres fix steps:



史诗级更新v1.2 -> v1.3:

  1. 增加更多细节的同时减少了污染

  2. 更好的背景,更好的背景,衣服,四肢逻辑

  3. 更好的光影效果

  4. 稍稍增加了对比度,画面看起来更加通透

可以参考下面的对比图

Huge improvements from v1.2 -> v1.3:

  1. More details while reducing unwanted elements.

  2. Improved background details, background, cloth, limbs logics.

  3. Improved lighting and shadows

  4. Slightly increased contrast for better "clarity".

Check the compares below:


这是我日常使用的融合模型。它有:

  1. 非常好的光影

  2. 比较好的手部

  3. 简单的用色,在使用画风LoRA的时候不会冲突、产生杂乱的颜色

  4. 非常少的污染,尽量减少提示词中没有的物品

我认为基础模型需要简单、少污染、会画各种概念。这样使用人物、画风LoRA的时候的可以更加好地还原。

建议负面:

(worst quality, low quality:1.4)

This is my everyday use model. it is aimed to:

  1. Getting great light effects

  2. Improving hands

  3. Being simple in coloring, so that it won't conflicts with style LoRAs

  4. Minimize objects that are not in prompts

I believe a good model should be simple, accurate, knowing many concepts. These are keys to make other LoRAs work well.

Suggested negative prompts:

(worst quality, low quality:1.4)