1 paper · 1 filter
Wenxuan Miao, Haosong Liu, Weiming Hu +9
Video diffusion transformers (vDiTs) generate high quality video but introduce extremely high compute cost due to the long diffusion timesteps and self attention computation. As di…