Building SameMuse: a private AI Clothes Changer for virtual try-on

I built SameMuse to solve a common online-shopping problem: a garment can look convincing on a store model while still being difficult to picture on your own photo. SameMuse is a private AI Clothes Changer that combines a personal photo with a garment image, screenshot, or public product URL to generate a virtual try-on preview. It helps compare silhouette, color, proportion, and styling without presenting the result as a sizing guarantee. The app uses Next.js, React, TypeScript, Tailwind CSS, PostgreSQL, Drizzle ORM, AWS S3 presigned uploads, and Sharp image processing. A key challenge was accepting realistic shopping references instead of only clean catalog images. Privacy was another constraint: photos, garment references, and generated results are deleted after seven days and are not used to train the model. The workflow stays simple: upload a photo, add a clothing reference, generate the preview, and compare it with the retailer's product information.