nilezt/CatVTON ? reverse-engineered prompt

Build me a virtual try on app for clothing photos.

I want to upload a person image and a garment image, then get back a realistic result of the person wearing that item. Please make it simple to run locally, with a small web app interface where I can choose the inputs, generate the try on image, and save the result. It should work well on a normal GPU and feel fast enough to demo.

If the project needs setup steps, include them, and make sure the first run can download whatever model files it needs automatically. I also want an inference script I can run from the command line for testing on a dataset, plus a basic evaluation script if that’s already part of the project.

Keep the experience polished, so someone can open the app, upload images, click generate, and see the output without much fuss. If you need current install or usage details, look them up online.