100 photos of a treasure chest with known camera poses go in. Thousands of 3D Gaussians come out, optimized until they reproduce every photo.