This is a 3D-aware diffusion-based model for generating diverse and "plausible" novel views from a single input image, incorporating geometry priors (assumptions about the geometric properties of objects) to build the world from the image.

→ View original post on X — @aibreakfast