Demult vs blend modes: which is more efficient?

I was about to send this question to the devs, but didn't want to add a support ticket for a non-support question.

Does anyone know if there is a difference in processing time between using a blend mode like Add or Screen compared to using the Demult effect? For a project I'm working on, I'm making lots of comps that are white elements on black backgrounds. Up until now I've just been going with my old standby and adding Demult to everything to knock out the black when layering them on top of the main footage.  However, I started to wonder if using a blend mode like Add or Screen (which effectively does the same thing) would make any difference in terms of processing time. In short, is one method more efficient than the other? (I'd run a test myself, but I'm on a time crunch to hit a certain milestone on this project, so I figured I'd toss this out for general discussion.)


  • NormanPCN
    NormanPCN Posts: 4,088 Enthusiast
    edited November 2018

    I would guess that Add/Screen would be faster than Demult+Normal. My reasoning is that the composite operation is always done regardless. Even Normal. Add/Screen are simple math ops and should execute similarly to Normal. Demult is an additional operation added to the dataflow. 

    That said, I don't think any appreciable compute time is saved. All these funcs are pretty fast. It would be hard for us users to test them since their overhead is soo low. Other overhead in the dataflow is likely to dominate with these simple operations. We users cannot really separate that in a test.

  • Triem23
    Triem23 Posts: 20,238 Ambassador

    I would guess blend modes are a bit faster. 

    I slightly disagree with @NormanPCN here in that I think Demult is really a simplified Luma Key, not really a blend mode analog. But Demult has to look at the layer, key out the dark, then pass to the layer's blend mode, so it's an extra step. 

This discussion has been closed.