Spatially adaptive representation of facial meshes for face video super-resolution
BFRffusion [Chen et al., 2024] aims to restore realistic and faithful facial details using a pre-trained generative prior based on stable diffusion.









