1 paper
Yixiao Jing, Chaoyu Zhang, Zixuan Zhong +1
Semantic noise initialization has been reported to improve robustness and controllability in image diffusion models. Whether these gains transfer to text-to-video (T2V) generation…