crisp and thin smoke

[dannyODF]

Hey. Im on a project right now where I have to build a character made of cigarette smoke. Ive decided to do it in Houdini. After a few days of exploring the program, it seems to be the right choice.

How would you make smoke that is this crisp:

since its an entire character, volumetrics only seem to get me 70 percent of the way. I cant seem to get extremely crisp/thin pieces of smoke with volumetrics. So Ive been exploring using millions of particles. I could also break the character down into segments to get more out of volumetrics?

here is what I have so far:

http://www.youtube.com/watch?v=urw1eF2Rwlg

any ideas would be much appreciated, thanks,

Edited November 6, 2010 by dannyODF

[rdg]

any ideas would be much appreciated, thanks,

Have you tried the upres'ing workflow?

If you go with particles have a look at clusterThis[1]

Maybe use the fluids to advect particles?

[1]

[zglynn]

The only way I know to get really tight looking smoke is to use millions of particles. I did this test a while ago and got some decent results. I advected particles through a low res volumetric sim, and cached out the particles several times using a wedge rop on the random seed. Then I used a delayed load render to call in the cached bgeos at render time, and using mantra I rendered them as points. Haven't messed around with the cluster this tool, but sounds like you could incorporate that as well.

Good Luck,

Zach

[dannyODF]

thanks, that ink in water looks really nice. I watched peter quints tutorial on setting up millions of particles, so if I can get something out of that, it might be an option. I gave maya's 3d fluid container a try too... and it can handle high res simulations quite well. The problem I keep running into with Houdini's smoke solver is sim times. The upres solver does not seem to work in this case because I need a fairly high resolution base simulation to even get something to come out of thin emitter objects. And when I turn up the res on this base simulation, the sim time sky rockets....

thanks for showing me that test you did, it gives me hope

[Serg]

thanks, that ink in water looks really nice. I watched peter quints tutorial on setting up millions of particles, so if I can get something out of that, it might be an option. I gave maya's 3d fluid container a try too... and it can handle high res simulations quite well. The problem I keep running into with Houdini's smoke solver is sim times. The upres solver does not seem to work in this case because I need a fairly high resolution base simulation to even get something to come out of thin emitter objects. And when I turn up the res on this base simulation, the sim time sky rockets....

thanks for showing me that test you did, it gives me hope

Did this type of smoke before in H using particles + curl noise voppop (replace with advect by volume if necessary) + a "neat" post trick.

Basically to pull it off convincingly you need detail as sharp as a pixel, particles can do that. The problem is that you need astronomical numbers to fill out the thin sheets as you look at them perpendicularly, otherwise it will look very erm... particle'ish... You can try to use particle multiplying techniques like running the simulation several times till you end up with 10s or 100s of million particles to convincingly fill the gaps... very costly slow process.

IMO clustering geometry shader's dont work for this because they just plonk orbiting particles around your simulated particles, thus creating blobs/loss in resolution/thickness limit.

But, by using a kick ass 2D de-noiser such as Neat Video you can quite easily fill/blur the space between the particles whilst retaining the sharpness as you look at the sheets edge on and matte lines, this worked very well for me with relatively few particles (say a couple million).

S