Nano Banana 2 E-commerce Product Photography Composition: Prompt Template for Negative Space and Materials

Use nano-banana-2 to run an e-commerce product image workflow of "reference image + prompt": first generate a white-background product reference image with the production line as image_input, then let the main prompt control the camera language, lighting, and background to output main-image quality that can be listed directly. The composition approach is built around the practical constraints of e-commerce main images: the subject is centered slightly low, space is left above for copy, and the background uses a low-saturation gradient with soft ground contact shadows to avoid a rigid pure white; for materials, emphasize the distinguishability of brushed metal, matte plastic, and glass highlights so the model can render them accurately during semantic understanding. At the parameter level, use a 4:5 vertical format (suitable for most e-commerce listings), 4K resolution, and PNG output for easy subsequent re-cropping and cutout. Applicable scenarios include: rapid prototyping of large batches of product main images and detail-page supporting images, design exploration that requires leaving promotional text areas on the image, and batch output of multi-category products under a unified lighting language. The reference image slot (image_input) keeps one of the 14-image upper limit, indicating that this is a batch usage of "anchoring on the reference image and using the prompt as the variable" rather than text-to-image from scratch.