Bring your face into the frame
Five frames, seven lens tints. Turn your head, take a photo, order on WhatsApp.
The camera image stays on this phone. Nothing is recorded or uploaded.
Everything runs on this phone. MediaPipe tracks the face and head; a small face-parsing network finds your real ears and nose, so the temples rest on your ears and disappear behind them, and the nose hides the frame where it should. Frames render in 3D with three.js: lit from your room's light, reflections from your surroundings, shadows on your face, sized from your iris. Photos are re-rendered at full resolution when you take them. Product names and prices are examples.
3D models (CC-BY 4.0)
Font: Urbanist (SIL OFL). Libraries: three.js (MIT), MediaPipe Tasks and its face landmark model, LiteRT.js (Apache 2.0). Face parsing: BiSeNet weights by yakhyo (MIT), trained on CelebAMask-HQ (research dataset).