xcc3641/blinkface ? reverse-engineered prompt

Build me a small app that lets me use my camera to frame my face with both hands, then restyle just that face area in real time with AI. I want it to feel like a live preview, so when I hold the hand frame around my face it should update instantly and switch between different looks like anime, Ghibli, cyberpunk, and similar styles when I close the frame.

It should also support a simple text prompt mode and an image to image mode, so I can generate a styled image from a prompt or from a photo. Make it work from a browser, and have a backend that can run on a CUDA GPU. If you need to look up current docs online for the model or webcam permissions, go ahead. Please keep it easy to run locally and include a simple command line way to test generations too.