• 3.0解决了一些图片崩坏问题。

• 3.0 addresses some image crash issues.

• 2.5相比2.0提升了真实感，更加的立体，不添加艺术家标签的话会有一点糊，不喜欢这种风格的话可以继续使用2.0

• 2.5 Compared with 2.0, the realism has been improved, it is more three-dimensional, and it will be a little mushy if you don't add an artist label, so if you don't like this style, you can continue to use 2.0

noobaiXLvPred模型融合了0.5+0.6+0.65s，加入训练的Lora，出图蛮抽象的，由于是V-Pred模型所以只能用Forge 和 ComfyUI

noobaiXLvPred model integrates 0.5+0.6+0.65s, adding the trained Lora, the graph is quite abstract
Since it is a V-Pred model, you can only use Forge and ComfyUI

分辨率：832 x 1216 （推荐放大1.5倍），1024 x 1536，1040 x 1520，

采样方法 (Sampler)：Euler（只能用这个）

迭代步数 (Steps)：28

CFG：5

masterpiece,best quality,amazing quality,very aesthetic,absurdres,newest,impactful picture,offcial art,movie perspective,

worst quality,low quality,logo,text,watermark,signature,