AvatarArtist: Open-Domain 4D Avatarization

Figure 1. Gallery of the proposed AvatarArtist. Each row features several triplets, where the first column of each triplet is the source image. The subsequent two images in each triplet are results that follow the pose and expression of the driving image, as demonstrated in the bottom right corner of the first three columns. Specifically, our method is applicable to an open domain, encompassing a diverse range of categories including 3D cartoons, video game characters, sculptures, skulls, etc.