E-commerce-Grade Product Photography Quality: Single Reference Image to Commercial-Grade Output

Using an ordinary product photo as a reference, re-render it into a product image at the level of an e-commerce detail page through the image-to-image pipeline of muse-image-1.0. The idea is to break "product photography" down into multi-part prompts that a reasoning-based workflow can decompose: first lock down the subject and materials, then specify the background and surface, lighting and reflections, lens and focal length, and finally post-processing quality (clean negative space, no messy shadows, sharp edges). The model plans the layout before rendering and aligns each part of the multi-part description item by item, making it suitable for batch-converting the same reference image into product image variants in different styles. Structurally, only two types of input are needed: one reference image (carrying the product's shape and brand details) and one structured English prompt; the aspect ratio and output format can also be specified. It is suitable for drafting e-commerce main images, category page scene images, and ad creative variants; since image_urls accepts up to 10 reference images, it can also be extended to a combined "product image + style reference image" usage. Note: public materials do not disclose technical details about output resolution or the number of reference images that can be blended, so this work uses only the parameters explicitly declared in the schema (prompt, image_urls, aspect_ratio, output_format).