Martin, if you're asking whether 'get ray origin' would be used instead of 'get camera position', they are very similar, but in this case I thought 'get camera position' was more appropriate because this is purely about the camera, and we don't need to make it work for reflections or other secondary rays. But I think 'get ray origin' would work just as well to allow us to calculate a vector from the camera/ray origin to the surface (you'd still need to do 'get position' minus 'get ray origin'). But neither solve the problem. Neither 'get camera position' nor 'get ray origin' will allow you to account for the rotation of the camera. I could generate vectors between the camera and the surface, and look at the difference between two rendered frames to get a motion vector. But you need to find the motion in screen space, which depends on how the camera is rotated, otherwise your motion blur will be rotated incorrectly.