希区柯克式变焦，两个模型，一个背景变大，一个变小，变小那个的效果不明显，但角色的动态还不错；变大那个的效果明显，但角色的动态不好。本来想把两种表现方式训练在一个模型内，但因为技术不够，使得会互相污染。如：视频中有一个（）Dolly zoom, one background becomes larger, one becomes smaller. The effect of the smaller one is not obvious, but the dynamic of the character is okay; the effect of the larger one is obvious, but the dynamic of the character is not good.Originally wanted to put the two types of training in the same model, but not enough technology, can't separate them.Label the body when using

Dolly zoom  wan2.1  i2v

背景变小，The background becomes smaller

Files with SHA256: ae77d421ac28b6904c9dcb11a4bfc6089ce1a52b8850f5d0a8b69a98504158cf