More Than Generation: Unifying Generation and Depth Estimation via Text-to-Image Diffusion Models | Read Paper on Bytez