Type | |
Stats | 347 389 88 |
Reviews | (39) |
Published | Apr 5, 2024 |
Base Model | |
Hash | AutoV2 AA79CBBDAB |
来群里扯淡吧:188546326
based on Animagine XL 3.1
基于Animagine XL 3.1
The weight only uses 1, which can also be used as a painting style.
权重仅使用1,也可以当做一种画风使用
————————————————————————————————————————
1.Please refer to my picture for the quality word.
2.The cfg recommendation is 3.
3.Try to enter all the characteristics of the character (if you don’t know, you can transfer the cartoon picture to the tagger and reverse it with 0.2).
4.Try to delete some work names, such as using nahida instead of nahida. (genshin impact)
5.If the background is too messy or the image is gray, try adding prompt words such as photo, photo background, realistic etc. in Negative.
6.This is an early version,There are no training pictures, all rely on prompt words, so it is a bit unstable.If there are many people using it, we will optimize it later...
1.质量词看我的图片
2.cfg尽量在3
3.尽量使用人物全部特征,单独触发词不太够用,如果你不知道,找个角色图片丢到tagger插件里以0.2阈值反推
4.尽量不要带作品名,他们会包含画风信息,比如nahida (genshin impact),只使用nahida,而不要用后面的genshin impact
5.如果背景太乱,或者图片发灰,尝试在Negative里加入photo,photo background,realistic等提示词
6.初期版本,并没有训练图片,全靠提示词,所以有点不太稳定,如果用的人多,后续再优化吧……