Matte Object double shadow

[spacematt]

I know there are many threads on these topics, but the only one I found with this specific problem was 12 years ago.



The red/green surface is a matte object, the spheres are 'real'. You can see double shadows at the foot of the steps, presumably caused by the shader finding 'real' shadowcasting objects first, and not testing for a matte object closer to the light source. Can this be solved without comp workarounds?

Also, while I'm here, is matte object workflow going to get any better? It's really hard work.

[yolov]

Can you share that example so that we can have a look, please?
Could you elaborate a bit on what you envision for making the workflow better? We could definitely take your input for consideration.

[spacematt]

Here's the file used to do the above image. Let me know if you have trouble deciphering it. There are several render layers...

1. Render the green & red surface for use as the background plate
2. Render the incoming light only, to be used in item 3.
3. Read the bg image and the incoming light image as a BG/environment for matte object testing.


matte_shadow.ma.zip

In terms of improvements, it's a complex topic, which gets wordy very fast, but for me the central issue is that it takes too many different techniques to achieve the basic set of desirable features. Everyone has different needs of course, but I like to get this stuff 99% right in the render and only break out the AOVs for creative adjustment if deadline permits. I also like to be able to create a 1-click solution to share with teammates, but that isn't possible here.

For example getting the chrome ball to correctly reflect the green/red checker texture takes a series of steps on both the wrapper node and the global scene settings. To get the white ball to correctly receive red/green GI from the checker texture requires a different set of steps, requiring rendering out intermediate images and reading them back in to the node network. Getting the matte object to correctly exhibit GI from scene objects requires a set of steps I haven't even been able to figure out yet.

Looking at it another way, the scene attached requires bringing the background image into Vray in three different ways, and writing then reading an intermediate file. All of this could be done entirely within one material. So, in the example scene.

1. Read in the bg image as an image plane for basic functions.
2. Read in the bg image as a secondary environment override with a Vray env placement node for correct reflections on the balls
3. Read in the bg image as a maya camera projection to map to the diffuse colour of the wrapper's base material (vray env placement doesn't work for this) to get semi correct GI on the balls
4. Read in a pre-rendered lighting pass of the matte object and divide the bg image before feeding it into the above maya projection, to attempt to remove lighting from the bg plate and deliver a more correct diffuse map.

The wrapper material could take care of all of this, with the aid of a few extra options.

1. The wrapper material ought to know what the background image is, and how it's being projected through the camera. That could eliminate the need for the user to set up an secondary environment override projection. Just use the same projection for reflection/refractions.

2. The wrapper material ought to know the incoming light values from scene lighting, and know that the bg plate is pre-lit, and could do the necessary division in the shader, to generate an estimated diffuse map for the base material. It should also know to use this diffuse map for outgoing GI (the red/green GI hitting the white balls)

3. The wrapper material should be able to take the diffuse map generated in 2. and multiply it by GI on the matte object thereby avoiding the overly bright or flat GI pools which appear below scene objects.