This is a typical particle simulation process that I use for most particle shot. Although the idea is simple, it’s been pretty useful.
First, I make a particle simulation with small number of particles in Maya or Houdini and run the simulation and generate caches using command line MTOR or Mayabatch or Houdinibatch on the farm with different random seed, so every simulation has slightly different particle position, and I combine all caches in render-time and render everything together, so I render a lot more particles. Of course, there cannot be any interaction between each iteration, but I rarely need those kind of interaction in production environment.
I’ve been using Pixar Alfred or Tractor for the farm manager, and usually I run about a hundred iterations, so I can get a hundred times more particle in same simulation time. In other word, I get the same amount of particle a hundred times faster.
Also, when I render those particles, I usually use Ri-Filter which is one of the Renderman Plug-in API that provides pretty convenient access to Renderman procedures and its parameters.
For particle rendering, I wrote very simple Ri-Filter that modify point size and generate motion blur block with particle velocity, so I don’t need to re-simulate to change point size, and I don’t need to double the size of particle cache to get a motion blur. I can also change the length of motion blur at render-time.
For this example clip, I used 50 iteration, so about 1 million particle simulation was done in 3 min on the farm.
Continue in Making Ocean #4