LeftRefill: Filling Right Canvas based on Left Reference through Generalized Text-to-Image Diffusion Model | Read Paper on Bytez