A very interesting paper about pipeline optimization from Filmakademie Baden-Wuerttemberg. Using newer rendering techniques could help reduce environmental impact, but also reduce iteration time, which is a key factor in efficiency that can be used both for budget optimization and better artistic results. The work essentially consists in re-engineering the pipeline for better throughput, and as funny as it may sound removing useless things that usually come from legacy practices.
I’ve spent a lot of my time when I was doing #color consulting (which I still do every now and then) reverse engineering rendering and compositing pipelines to find where the problem came from, and in general we had in the node tree something that had been there forever, nobody knew who put it in the first place, combining layers rendered with weird options. As much as the “EXR movement” brought a lot of good things for flexibility, allowing to recombine multiple layers and benefitting from the floating point concept, a lot have forgotten why it was there in the first place : to reduce the risk of having to re-render, which back in the days was extremely expensive on SUN and Silicon Graphics workstations. We now have in a typical studio hundreds of (not so cheap) machines calculating hundreds of layers at rendering, then those layers tweaked in comp, sometimes for no good reason and the result processed again, and then going to color grading for a final rounding error. And I’m not even talking about people using #PBR for non photorealistic #CG bending the parameters to get a result they could get with much lighter setups.
The main problem really is that it’s usually considered too expensive to put some brains to think about the problem than doin’ it the good ol’ way, that’s also why so many “real time” transitions didn’t go too well…That’s why the approach in the article is interesting, they have put the effort to see what’s possible in saving, and that’s pretty good.
Obviously it won’t work for all projects, it brings some risks when nobody want some, but at least we know it’s possible.
Cedric
PS : it feels good to see my work for the Green Screen Interreg project cited!
Originally published on Splectrum in Nov 2024. You can continue that discussion on our Discord server