Learning Visual Generative Priors without Text | Read Paper on Bytez