hey Matt, thanks a lot for the detailed explanation and examples! that's exactly what I neeed. much appreciated.
2014-08-28 18:48 GMT-03:00 Matt Lind <[email protected]>: > Updated version: > > > > Vector_state à vector-Vector à Vector_Coordinate_Converter à > Vector-Scalar à Gradient à Surface > > (normal) (negate) (Vector, to Camera) (Dot > against -1,0,0) (scalar input,[-1…1]) > > > > > > This works in similar fashion to the method I submitted previously. The > main difference being the normal vector is dotted against the camera’s –X > axis to provide the directional information. > > > > If the viewed normal aligns with the camera’s –X axis, the dot product > will return -1 (red). > > If the viewed normal vector aligns with the camera (points to the camera), > the dot product will return 0 (cyan). > > If the viewed normal vector aligns with the camera’s local X axis, the dot > product will return 1 (yellow). > > > > The output of the dot product drives the gradient. Just make sure to use > Processing > Mixers > Gradient and not Texture > Gradient. If you want to > make the output more accurate, zero out the Y component of the normal > vector after it’s converted to camera space, but before doing the dot > product. Then normalize. That will eliminate the vertical component. > > > > > > You want to avoid using expressions in render trees as it can slow down > performance and cause other unwanted side effects. Mental ray does all > computations in ‘internal space’ which is largely undefined. Softimage did > it’s best to ensure that for lights and cameras, internal space is > equivalent to world space. So when the ‘vector_coordinate_converter’ node > is used above to convert the normal to camera space, what it’s doing is > converting the normal from world space to the local coordinate space of the > camera. However, it should not be assumed that internal space always maps > to world space, because it doesn’t. > > > > > > Matt > > > > > > > > > > > > *From:* Matt Lind > *Sent:* Thursday, August 28, 2014 11:35 AM > *To:* [email protected] > *Subject:* RE: Dot Product on Rendertree? > > > > You cannot get your desired result with only a dot product because in > order to differentiate between facing left and facing right, you need > directional information. A dot product only gives the cosine of the angle > between the vectors, not direction. > > > > Based on your description, I’m assuming you want to drive color based on > the surface normal relative to the camera. To do that is fairly simple. > > > > 1) Get a vector_state node and set it to ‘normal_vector’. This is the > surface normal at the intersection point. > > > > 2) Get a vector_vector node and plug the previous node into > vector_input1. Set the operation to ‘negate’. This will flip the normal > to point away from the camera (assuming it’s a front facing surface). > > > > 3) plug the result into a vector_coordinate_converter node, and set ‘type’ > to ‘vector’, then set ‘transform’ to ‘to Camera’. This will treat the > normal as an orientation vector and re-describe it in the camera’s > coordinate space. We do this to isolate the normal vector’s X-axis to make > it easy to determine if it’s facing left or right relative to the frame. > > > > 4) Plug the output of the vector_coordinate_conversion node into a > ‘vector_to_scalars’ node. This will give us access to the normal vector’s > X, Y, and Z components. > > > > 5) Plug the ‘X’ output of the vector_to_scalars node into the ‘input’ port > of a gradient node. In the gradient shader, go to the ‘input’ tab and set > active range to [-1…1] and set input type to ‘scalar input’. Finally, > adjust the color gradient markers so the left end is yellow, the middle is > cyan, and the right side is red – the opposite of the default arrangement. > If you do not want to futz with the color markers, then invert the ‘active > range’ on the input to be [1…-1]. I don’t recommend inverting as it’s > non-standard math and while it’s perfectly fine for this simple prototype, > it’ll cause headaches in a production pipeline as it’ll be a source of hard > to find bugs from stupid logic. > > > > > > > > Matt > > > > > > > > > > *From:* [email protected] [ > mailto:[email protected] > <[email protected]>] *On Behalf Of *Fabricio Chamon > *Sent:* Thursday, August 28, 2014 8:20 AM > *To:* [email protected] > *Subject:* Dot Product on Rendertree? > > > > Hello fellow survivors! > > > > I'm trying to gradient shade a simple grid, using the dot product of > surface normal vs. camera position: > > > > > > basic idea is: > > grid facing camera > cyan > > grid facing left > red > > grid facing right > yellow > > > > the above setup is not working... any suggestions ? > > > > thanks >
