Announcement

Collapse
No announcement yet.

General VFX company workflow - Mattes and ID passes in main render or separate?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • General VFX company workflow - Mattes and ID passes in main render or separate?

    Heya Folks,

    Just thinking about some of our setup at the minute and looking at optimizing render times. I'm working at film res on a lot of big vehicle stuff with complex materials, and I'm noticing the extra load that having a dirt map puts on to the render times. I was going to spend a bit of time seeing could I optimize things to get the best combination of render time and acceptable quality, but on simple tests I'm noticing that the AO pass adds maybe a 20% penalty on top of the render time. With the unified sampling methods being a careful balancing act do most people just break out the matte and utility parts into a separate render with the AO stuck in here instead? Anyone working on film vfx care to share their experience?

    Cheers!

    John

  • #2
    what I usually do in my setups is implement ao as a utility pass, but leave it off, and only render it on demand from comp, since as you mention it does add to the render time. Generally speaking we split all utility passes into separate render, for things like different float points, some utility needs to be 32 bit, but rgba beauty does not have to be, and at present there is no option to specify which pass gets which float. Also, if you have an error in a utility pass you don't have to re-render the beauty which could be of some benefit. Of course there can also be more utility passes then beauty, for example if you have to integrate with live plate, you would need to render contact shadow passes, and possibly other passes for the ground, like motion vectors, zdepth or whatever else.
    In a larger scale production we typically script all of the pass output, so a lighting artist(s) don't have to worry about getting the right amount of passes to comp, that's taken care of by look dev artists. All you do is press a button - beauty/utility and you get your scene ready for rendering, so everything is consistent across the board (huge time/life saver).
    Dmitry Vinnik
    Silhouette Images Inc.
    ShowReel:
    https://www.youtube.com/watch?v=qxSJlvSwAhA
    https://www.linkedin.com/in/dmitry-v...-identity-name

    Comment


    • #3
      Yep I've got something nearly the same aside from some passes relying on the vray extra tex and selection sets which unfortunately don't have script access just yet in max but all the render presets and naming are consistent. Which utilities are 32 bit? Z depth and the likes?

      Comment


      • #4
        sometimes, for extended comps (and Im not a pro compositor to know how exactly they used them) we required to give them 32 bit data. To get proper info from passes like world/object position, especially if the object is far from origin, a 16 bit half may look correct but surely won't work in all cases, 32 bit will. Zdepth, for example there are some studios that develop specific tools where you just had drag n drop zdepth setup, where its all world based, using world position as well as 32 bit zdepth to achieve that.

        I think it also works a bit differently in max then in maya, where in max we had to construct the passes on the fly, scene by scene. In maya we just make an empty scene which has all the passes, and that gets referenced into all other scenes, so there is no need to delete/rebuild them.
        Dmitry Vinnik
        Silhouette Images Inc.
        ShowReel:
        https://www.youtube.com/watch?v=qxSJlvSwAhA
        https://www.linkedin.com/in/dmitry-v...-identity-name

        Comment

        Working...
        X