SideFX - Pyro COP Experiments
This film cannot be played here. Open the video file
This film cannot be played here. Open the video file
This film cannot be played here. Open the video file
This film cannot be played here. Open the video file
During my internship at SideFX I spent much of my time on the new Pyro solver in Copernicus. Unlike the classic DOP Pyro, it is written in OpenCL and runs sparsely on the GPU, inside the same image network as everything else in COPs: a source can be built there, or brought in from SOPs, and the simulation runs live while it is being changed.
Because it is so fast, I could treat smoke the way I would treat an image. The same small explosion took about five seconds in COPs against thirty-eight in DOPs; a low-resolution version of the colourful explosion cooked in about nine seconds against five minutes, and the full version reached about a hundred and twenty million active voxels. Blur, blend and max operations from the image toolset shape the motion directly: blur on velocity reads as viscosity, and blending velocity by length rather than per axis keeps the strongest push in any direction.
The fire tornado took about a month. A custom curve force shapes the particle source into a tornado profile, the smoke is held together with axis suction, and the fire is not simulated at all: it is sculpted from noise and relative density, then baked into the smoke so it follows the same motion. I presented the work at Houdini Equinox in Toronto, and the setups are available in the SideFX Content Library.
Credits
- Client
- SideFX
- Role
- Marketing FX Artist (Intern)
- Tools
- Houdini, Copernicus, and Karma
Team
- All-round support
- Fianna Wong
- Pyro support
- Attila Torok
- COP Pyro developer
- Jeff Lait
- R&D help
- Jeff Guttin
- R&D help
- Nick Culver



























