Trouble retiming a smoke sim.

[ben_hamburg]

hey there,

recently posted this question in the facebook group, but I´m still having trouble with teh retiming of a smoke sim.

https://youtu.be/RMGGcaYbPdE

As you can see I get some ugly jittering in the retimed version.
I used the general purpose method for the retiming and amped the resolution quite a bit (from 25mill to 200 mill voxels).

There is some confusion in the manual about the retiming in the tips and tricks chapter:

"Do not change any settings in the Dynamics rollout if you want to get a correct Time Bend Resimulation. You can experiment, of course, but results will typically have flickering or distorted playback."

and some lines later:

"Note that both methods work best for simulations with Steps Per Frame set to a value of 1;"

As my main sim was simmed with SPF 2, I only remembered the latter note, so I resimmed the retimed version with SPF lowered to 1.

I´m not sure now, if the jittering comes from that and how to best procede now:

1. resim the retimed version with SPF 2.
OR
2. resim the base sim with SPF of 1 and then resim the retimed version with SPF of 1 as well?

As you can see the forces are quite fast in this, so I´m wondering if I would get velocity artefacts with SPF set to 1.

[Svetlin.Nikolov]

Yup, if your original sim is simulated with SPF > 1, then it's very likely that anything you do will result in a messed up resimulation. Resimulation depends on the velocities from the original simulation, and since there is no data for what the velocities were in the substeps, resimulation has no chance of replicating the original sim if it uses more than 1 step per frame...

[ben_hamburg]

Dang.
But which would lead to cleaner results: Resimming the retimed version now with SPF 2, or resimming the base AND the retimed version with SPF 1 instead?

[Svetlin.Nikolov]

I wouldn't be able to guess, but since it's not looking good now, I'd suggest you:
- 1. Resimulate with 2 steps per frame and NO resolution amplification; if it's still not looking good, then:
- 2. Copy the scene and simulate the original with 1 step per frame; then resimulate with NO resolution amplification; if it's not looking good, then the reason might be completely different so we'd know we should be looking elsewhere
- 3. Otherwise, if the sim looks good now, amp up the res

This way you'd isolate any factors one at a time and know for sure which one affects negatively the sim.

[ben_hamburg]

Yeah, those are some good tips...

I´ll ignore the first one though, as I wanna change some stuff with the vortex force anyways.

On a side note:
As I was crossposting on facebook as welll...A lot of times facebook is more helpful nowadays, as people just frequent it more often.
But I hate finding info there, so with anything that might come up more often, I´m also posting it here...

[Svetlin.Nikolov]

Both ways are fine with me, as long as the info can be looked up later by someone else