Animating General Image with Large Visual Motion Model
Dengsheng Chen, Xiaoming Wei, Xiaolin Wei
We present the pioneering Large Visual Motion Model (LVMM) meticulously engineered to analyze the intrinsic dynamics encapsulated within real-world imagery. Our model fortified with a wealth of prior knowledge extracted from billions of image pairs demonstrates promising results in predicting a diverse spectrum of scene dynamics. As a result it can infuse any generic image with authentic dynamic effects enhancing its visual allure.
