(转载自C站white2023)
历尽千辛万苦，终于出来一稿像样的DB
虽然还是不满意，不过作为V1版本应该可以发了吧
虽然写着MIX，但是基本上融的都是我自己的lora，所以写个TRAINED不过分吧
因为不知名原因，高清重绘开高了容易出现图片质量下降，测试下来开0.35比较合适，或许你们有更好的使用方法可以留言告诉我
建议使用动态CFG插件（CFG Scale Fix），会获得更好的效果
还是和之前的lora一样，使用更精确的提示词会获得更好的效果，当然你们也可以参考我的TAG
做这个模型的初衷是为了尝试打破那千篇一律的AI感，做到没做到，答案在你们
我自己也说不好，玩的开心朋友们。如果喜欢，请给我返图，期待看到你们的作品
商务合作请联系B站：咒术师老白
After going through countless hardships, a decent DB has finally been produced
Although I am still not satisfied, as the V1 version, I should be able to send it now
Although it says MIX, it basically blends my own Lora, so it's not too much to write 'TRAINED'
Due to unknown reasons, high-definition redrawing with a higher opening may lead to a decrease in image quality. After testing, opening 0.35 is more suitable. Perhaps you have a better way to use it, please leave a message and let me know
Suggest using the dynamic CFG plugin (CFG Scale Fix) to achieve better results
Just like the previous Lora, using more precise prompt words will achieve better results. Of course, you can also refer to my TAG
The original intention of making this model is to try to break the stereotype of AI, but if it is not achieved, the answer lies with you
I can't say well myself, have a good time, friends. If you like it, please return the pictures to me. Looking forward to seeing your works
For business cooperation, please contact Bilibili: 咒术师老白
