A small test with some advection at rendertime (or gridless advection if you will). I start off with a simple pyro sim, and then I take the density field and advect that with the velocity field of the same frame for 8 steps at rendertime, for every frame of the sequence. The result is very detailed since you aren't limited by voxels, and you also get rid of a lot of artifacts related to the same.
For the second example I simply added some curl noise to the velocity field post-sim and wrote that to disk.
Loading more stuff…
Hmm…it looks like things are taking a while to load. Try again?