Google Virtual Try-On Portrait: A Dual-Prompt Demo of Realism and Artistic Stylization

A try-on demo entry for the google-virtual-try-on image pipeline. This pipeline takes two input images: a full-body photo of a person and a product image of a garment, and outputs a single try-on image of that person wearing that garment. This piece builds a minimal runnable input from reusable production assets: the person reference image is a front-facing standing full-body portrait under natural light, making it easy for the model to reliably locate body proportions and the clothing coverage area; the garment reference image is a flat-lay product shot on a solid background, reducing background interference. The two prompts cover two distinct use cases, "realistic pre-purchase reference" and "artistic style reference": the former emphasizes the naturalness of fabric folds, drape, fit, and lighting, suiting e-commerce detail pages and shopping-decision scenarios; the latter preserves the try-on structure while strengthening stylized texture, suiting outfit inspiration and social sharing. The input strictly follows the author's pipeline schema: the required fields person_image_url and product_image_url are both generated and bound from assets, and num_images is fixed at 1. This schema does not declare a prompt field, so the underlying submission uses only the two images; the top-level prompt of this piece is used solely to present the creative intent on the detail page. Developers can use this to judge the input form and the expected image composition of this model family on virtual try-on tasks.