Would love thoughts from people with image generation experience: we're trying to generate photos of blank aluminum cans using stable diffusion (sdxl) and we're trying to overlay those cans with a label. Would the "correct" method be to create a normal/depth map of the generated image, mask the product, then overlay the label onto the masked normal/depth map? We want to make sure the label correctly aligns with the creases/curvature of the can