Originally Posted by derOesi
so ..uh... you guys dont work 24x 1h shifts on a single workstation like the rest of us? could you share some insight on this new technology used by digic(2012) and!! JurassicPark (1995)?
sorry couldnt resist
Well it isn't new technology. It is a philosophy of working and organizing folder structures for the production.
There was a very old "Making of" by Digic that explained for example how their folder structure was simply divided between Scenes and Assets and not a literal forest of folders. We felt that was a good philosophy because you can spend a long time designing folder structures and Digic's concept of just starting with two was very good.
The other thing we saw in that old Digic workflow was the low-poly/hi-poly animatic-direct-to-finals style of production. We took this to mean that if we could find the correct point during modeling, we could actually start importing the actor into scenes so that animators could work with the character even while modeling on the actor's hi-res mesh was still ongoing. This made things parallel.
We took the same philosophy looking at everything else: Music, Audio, animation, modeling, texturing. We tried to launch as many things at the same time as possible. So it's a parallel philosophy.
Jurassic Park, and this is my personal opinion. Jurassic Park is the perfect project case study for "first time endeavors". No one knew yet how to do any CGI dinosaurs in Jurassic Park and the project had gone down one direction before shifting to another. Not to mention the same case study featured an episode of disaster recovery and you have the perfect film case.
Jurassic Park was very well managed in the sense that the manpower came in a JIT (Just-In-Time) sort of way and there was very heavy emphasis on testing.
So we took that concept and some of Jurassic Park's methods of controlling risk, testing VFX, and introducing artists in the process of the film, combined with a very Parallel system, together with the concept of working in small teams that could each produce output.
We also shunned doing Beauty Renders at the very end and instead doing them for each sequence and running these final colors independent of any rendering farm. The concept was that a farm probably doesn't matter in a system where more than one scene is rendering at once and each scene is optimized for single PC's in testing.
So it's not technology.