FMX 2013 - The "unfilmable" Life of Pi
Date: 23.Apr 2013 | Reading Time: 1 minute, 45 seconds
Opening with a joke about the botched Oscar ceremony, this is promising to be a good yet sad talk. Rhythm & Hues went from 1000 artists to a small fraction of that recently.
Why was Life of Pi “unfilmable”
Three simple reasons, combined making for a perfect storm:
- kids (the actor playing Pi while technically not a kid played his first role and could not swim)
Ang Lee did some hands on research of how a life boat or raft behaves in the ocean and how water and waves behave. Authenticity was paramount.
A pool with a wave machine was built in addition to an extensive set of 3D shaders mimicking the pool water. The water had to blend with the CG seamlessly.
- custom, physically plausible water shader
- a set of five different water noises to layer and mix the water as needed
- to avoid tiling a set of noise patterns was used to multiply the effects of the water noise parameters
- once water was dialed in, it was locked and 3D artists couldn’t change it and needed to adjust the scene to get the desired effect
- and of course, physical accuracy went out the window as soon as we reach the comp stage or when the director wants the water different then the physical simulation says it should look
About 140 different extremely high–res HDRi skies were shot and and artists could pick and choose crop outs
- one giant single Banyan tree system with about 6 billion polys per frame
- up to 45000 meerkats in one shot done with Massive
- up to 10000 in the shot
- basically only Pi is real, the island is CG
Richard Parker, the tiger
- subsurface on the fur
- new muscle system
- not anthropomorphized at all
- based on a real tiger, “King”, from France, he appears in only 23 shots
muscles simulation → subcutaneous skin layer → epidermis layer that slides over the top → covered by fur
All that leads to the realistic wiggling, bouncing and sliding of ski that makes for a realistic animal.
10 million strands of hair lit with area lights, subsurface scattering