It’s more of a reconstruction technique than rendering technique. But the power is the 3D differentiable renderer. That means we can optimize an output image to look exactly as we want - given sufficient input data. If you want to learn more take a look at differentiable rendering and 3D multiview reconstruction.