compositing question

[royter]

i want to render 2 layers: one for my box in the middle and one for the rest of teh scene.
first, i render everything in one layer (first image), *here i freez FG
then, everything but the box (primary visibilty off)
then, just the box (with all the rest primary visibilty checked off).
then, an alpha mask with surface shaders.
(render settings 0 2 ,0.01, lanczos, premultiply cheked OFF, PNG format)



The problem is that in my final composed image in PS, i dont get the same result as the first render where everything was rendered in one pass. i get a kind of white line on the contour of my box:
-Why is that?
- Is there anything wrong in my workflow?

http://sor.typepad.com/.a/6a00d8350600cb53ef0105361f264f970b-pi

[MasterZap]

Sounds like a premultiplication issue. Are you interpreting your alpha as premultiplied in your compositing software?

/Z

[royter]

i don't know how to interpret my alpha as premultiplied in PS.
i am rendering the alpha mask in maya with no transparency (only black/white), then i paste this mask in my alpha channel, drag that channel to the selection circle and delete the unwanted portion of the cube.

[svv3d]

"Compositing Premultiplied 3D CG in Photoshop"
http://www.digitalartform.com/archives/2005/10/compositing_pre.html

[royter]

actually i noticed that even in the first step (before applying the alpha),
the rendered brick alone (over the background layer) doesnt' look like the brick rendered with the background in one pass. Shouldn't they look exactly identical given the fact that i only modified the primary visibility of the background objects and the brick simultaniously and freezed my FG solution?

http://sor.typepad.com/.a/6a00d8350600cb53ef01053629bd7d970c-pi