Level of detail in an age of GI: rethinking crowd rendering

In the past few years, the feature film animation and vfx community have made a marked move toward single pass, physically based, global illumination (GI) rendering. Improvements in hardware, software, and algorithms, have allowed studios simplify their lighting setups and produce more photorealistic imagery. But woe to the early adopters who thought they could render massive scenes using the same techniques that allowed rasterizers to churn through near unlimited complexity. Anecdotes across studios, using a variety of pipelines/renderers, suggest that 100+ core hour renders have not been uncommon, leading to limited iteration, rushed hardware purchases, and panicky producers. Pixar had a brush with such panic on the film Monsters University, where rendering crowds of highly detailed monsters, revealed the challenges of scale inherent to the in core nature of physically plausible rendering. This launched an effort to re-evaluate the various level of detail techniques used at the studio and explore new variations more amenable to GI.