1 paper · 1 filter
Orest Kupyn, Théo Uscidda, Marta Tintore Gazulla +3
Video generation models have advanced significantly through the latent diffusion transformers trained with rectified flow techniques. Yet these models still struggle with geometric…