The training set is all from the young Chtholly in the early days, so you cannot use this model to draw the later, red-haired character form.
The bottom model is anything3.0, after testing, I personally feel that it is in a better state in anything3.0.
Under the model of anything3.0, the better parameter is between 0.6 - 0.8
Under the counterfeit2.5 model, it is necessary to slightly increase the parameters introduced by lora. It is recommended that > 0.7 will bring better results. (need to add vae, otherwise the overall picture will appear a bit pale)
This model does not fit the chilloutmix model well, so use it with caution
Can be triggered without a keyword, or through the chtholly keyword.
训练集全部来自于,比较早期时候,年轻状态的珂朵莉,所以你并不能通过该模型绘制出后期,红头发的人物形态。
底模为 anything3.0 ,经过测试,个人感觉在 anything3.0中有比较好的状态。
在 anything3.0 的模型下,较好的参数为 0.6 - 0.8 之间
在 counterfeit2.5 的模型下,需要略微调高 lora 引入的参数,建议 > 0.7 会带来比较好的效果。(需要加vae,不然整体画面会显得有点淡色)
该模型对 chilloutmix 模型适配不好,谨慎使用
可以不使用关键字触发,也可以通过 chtholly 关键词来触发。
use anything3.0 as a basic modal
Go ahead and upload yours!
Your query returned no results – please try removing some filters or trying a different term.