although you can't see it here the original particle positions are set according to a 3d scan of someone's face. I view that model in a renderer, output a depth buffer, convert it to linear depth, then sample all those pixels in a compute shader with one particle per pixel.
how isit done?
dx11 particles, modified using 2d textures with a few post process stages (dof,shadows,lomo) in deferred rendering style
although you can't see it here the original particle positions are set according to a 3d scan of someone's face. I view that model in a renderer, output a depth buffer, convert it to linear depth, then sample all those pixels in a compute shader with one particle per pixel.