“Chinese video generation models like C-Dance 2 are outperforming U.S. counterparts, likely due to their ability to train on a wider range of data.”