NeRF represents a scene as a continuous function, usually neural network that takes a 3D position and view direction and outputs things like color and density

NeRF shoots rays through the scene and sample many points along each ray, blending them with volume rendering

Downside is that it’s often slow to render