debugged, but I personally like the current test and picture production, so I will share it with you. Everyone is welcome to produce pictures and ask for a picture return. Then I will adjust the final effect of the model based on the picture return. Trigger word: a image in style of XSF_DAI. Then create some descriptive words for the model, such as luminous trees, stars, etc., which seems pretty good for text generation at the moment, but it is not sure whether it is really stable.

Files with SHA256: ec9b89ee6c201f4199ca4c64bd823e5c4c43fa4ea017a442b7f61fc59a5f01e3