I trying to speed up cycles 8k render (even completely empty space took significant time to render..idkw!) by rendering border (region) and fill rest of transparent space with solid color.

Result I want achieve (render just border (region) and in composition fill rest with color):

I want this ;)

But I have this result which is wrong..

I got this.

Simple composition works just fine for whole render view. Composition

Mayby I'm missing something - so question is simple - Is there any way to render border (region) and then fill outer area in composition?

Edit: After testing it's possible to have good result just by changing window in blender from 3d view to node editor.. But scene is scripted so this solution doesn't work for me :(

PS: I cannot use image as background - it has to be generated node as RGB or similar because is part of bigger project.

  • 3
    $\begingroup$ Could you spend a few more words about what exactly you are not able to achieve? $\endgroup$
    – Carlo
    Mar 3, 2018 at 22:39
  • $\begingroup$ Hi, sorry for my poor explanation. I'd like to render just region (border) and fill transparency with color to achieve figure 1. Now with composition (figure 3) if I render border I got figure 2. $\endgroup$
    – Destrosvet
    Mar 3, 2018 at 22:49
  • $\begingroup$ It should work with your set up, but I tried myself and it isn't. The cropping of the image is kept beacuse the nodetree is not being correctly updated. Moving any slider, even if it belongs to an un-related node, forces the update, and all works fine. I think it may be an unfixed bug: the way you are doing is in my opinion correct. $\endgroup$
    – Carlo
    Mar 3, 2018 at 23:08
  • $\begingroup$ Thank you Carlo for assurance that is correct way - it helps. I went mad yesterday. Because source scene is scripted, I'll try to update composition by script and hope for good result ;) So should I report this as a bug? I'm using 2.79a, didn't tested on other versions.. $\endgroup$
    – Destrosvet
    Mar 4, 2018 at 6:32
  • $\begingroup$ Same result in 2.78c :/ $\endgroup$
    – Destrosvet
    Mar 4, 2018 at 9:07


You must log in to answer this question.