Repost: https://huggingface.co/lkzd7/Z-Image-LoraSet-NSFW

This is a LoRA specifically optimized for comparison views. Although ZIT could originally render multiple views, there were still some issues:

1. Each sub-image required a detailed description.

2. Certain descriptions meant for a single sub-image might inevitably appear across all sub-images, such as exposing breasts.

This LoRA only requires a simple description of the initial outfit, and it can automatically generate outfits for the other sub-images.

It is most stable when the number of sub-images is 2, and 3 sub-images also perform well. When there are more than 3 sub-images, it becomes very unstable (the exposure order may get mixed up).

This is a LoRA specifically optimized for comparison views. Although ZIT could originally render multi-views, there were some issues:

1. Each sub-image needs to be described in detail.

2. Certain descriptions targeted at a single sub-image may inevitably appear in all sub-images, such as showing breasts, etc.

This LoRA only needs a simple description of the initial outfit to automatically generate outfits for other sub-images.

The most stable number of sub-images is 2, and 3 also performs well. When the number of sub-images exceeds 3, it becomes very unstable (the degree of exposure may get out of order).