QUOTE
Let's say I render the box alone, and then the torus with a black matte material assigned to the box.
Now if I comp these two together, putting the torus (with black where the box is) over the box, there is a same problem.
Ok i did this no problem. No gap. Also with particles no gap.
But yes this is classic sign of false interpretation.
QUOTE
hese are coming from maya, so they are premultiplied.
Ok, see maya doesn't HAVE to be premutiplied there's a option in render settings to choose what you want,and is the premult color in AE right. Also chek the bg color is set to black.
o oods are the other is premultiplied while the other might not be.
QUOTE
With no antialiasing, there are no transparent pixels on the edges so no problem.
Yes offcourse no alpha interpretation error.
But yes it might be aa issue among other things, it would happen If you would choose to use mentalray and use a higher order filter, use box (in maya software dont use multipixel filtering). But then if you use a nurbs sphere youd still be off a bit since the tessellation is different in the renderers.
Also chek that the famebuffers are linear, if you adjust gamma the aa isnt going to match.
Then theres the small fact that mentalray isn't guaranteed to have same rounding as maya. this is why you can render particles in mentalray.
Personally i dont do it thisway, as theres a better option of using zcomp inside maya hardware renderer. Or in shake (but good luck with this in AE)
QUOTE
And I checked the correct option in after effects.
Please re chek. I have had this duiscussion about 50 times and it was still wong because after effects didnt do whatyou tough you told it. Because in few cases the global was set and in other not.