建议搭配我的模型使用

lora模型地址： https://www.liblib.art/modelinfo/d9675e37370e493ab8bf52046827a2b0?from=personal_page&versionUuid=69e5d34a4ffb4e5e8b3212579a7d6fba

unet模型地址： https://www.liblib.art/modelinfo/25236ae1ea604218b5e8880ceab89fc8?from=personal_page&versionUuid=5fe9292ab49743dbabccf3056613b0d3

反推可以换成Joycaption效果更好，理论上讲用Joycaption，传什么图片，最终都会出来与传的图相似的图

It is recommended to try it with my model.

Lora model address:

https://www.liblib.art/modelinfo/d9675e37370e493ab8bf52046827a2b0?from=personal_page&versionUuid=69e5d34a4ffb4e5e8b3212579a7d6fba.

Unet model address:

https://www.liblib.art/modelinfo/25236ae1ea604218b5e8880ceab89fc8?from=personal_page&versionUuid=5fe9292ab49743dbabccf3056613b0d3.

For inversion, replacing it with Joycaption will have better results. In theory, when using Joycaption, no matter what image is uploaded, an image similar to the uploaded image will eventually be produced.