1 paper · 1 filter
Zhaochong An, Orest Kupyn, Théo Uscidda +5
Large-scale video diffusion models achieve impressive visual quality, yet often fail to preserve geometric consistency. Prior approaches improve consistency either by augmenting th…