agentfx

original 200x150x100 
uprezed 2x
delta 4
Str 2
Octaves 2
Steps 15

in this second test, (i'll sort of repeat myself from the vimeo description) I wanted to push the values too far (in delta as this test was built with version 1.27 and I say that half joking since peter made 2 revisions to version 1.27 w/o versioning)
Anyway so I wanted to push the delta further than I was to really see what happens. Its still cool but it gets a little rough in the movements compared to the original 

check out this fire reference
fire starts at 1:16, but my point in showing it is to get a sense of how high we really can/need to push the delta to achieve realism. 
--Dave
Quote 0 0
Cainor
Quite the amazing test. Currently I'm testing wavelets on just pure smoke with lots of movement. But its very noticeable that a fire or explosion simulation with wavelet's even on a minor setting, add that detail that brings it closer to realism.

Great work, keep them coming

Quote 0 0
pshipkov
Very nice example there.
Yes, pushing the WT delta value is the key to get the nice billowing that makes things to look real(ish).

And this fire reference video is really amazing. I cannot stop watching it.

Quote 0 0
JeremyR
nice reference, great tests also.
Quote 0 0
ourouk
Hi everyone,

Need some help on a case : I tried on many examples but I have often a consistency problem of the wavelets over time.

See the video :

password : soup
The video is not public.

Resolution : 3
Advect Density and Temperature ON
Method : noise advection ( MacCormack )
Delta : 0.5
Strength : 2
Octaves : 2
Max substeps : 5
Others params : default values

Any idea ? too much strength ? too much Octaves ? Delta ?
thanks

Quote 0 0
ourouk
Reducing delta to .25 improve the result but the look of the smoke is not as good as with delta 0.5

I think I understood the purpose of delta, but I cannot understand why increasing it to .5 create "flickering" wavelets ?
Quote 0 0
pshipkov
Delta is a value that defines the distance at which density/temperature get advected along the velocity vectors. So if you "project" the density too far from it's origin things may very well start getting funky. Think about it and you will see what i mean.

Quote 0 0
agentfx
new example 
--Dave
Quote 0 0
agentfx
there will be another video after that with a different comp treatment. and you can download the source footage I believe, w/o compression. I'll check into that but it should be a feature of vimeo plus that I have. 
--Dave
Quote 0 0
JeremyR
agentfx wrote:
new example 


I can see some awesome realism between dark and light half way through the animation. Its looking a bit unreal in the beginning though but obviously its a WIP. Nice.
Quote 0 0
agentfx
well the issue with the beginning is, I wanted to make it slow motion but that requires a slow start, which I could either turn the wavelet way down at the beginning and key it up (hoping to get that right with each cache iteration)
The other trouble with this example is it will never look super great, unless the original sim looks super great. I need to go back a few steps to go forward, and I will, just I'm getting tired of my computer being frozen siming or caching. I'm thinking its as far as I want to push this sim. I want to fix the original sim and get that closer to looking good on its own. Its 120x so its got a lot of room for more detail, before I uprez/wavelet.
I'll save the shader settings. But I've got to figure out how to shade smoke and fire with only one opacity graph. 
Also I'm not following Peters instructions on this to sort of normalize the velocities on the wavelet with the mult settings. 
--Dave
Quote 0 0
JeremyR
Its just the speed of the turbulence near the source that bothers me, not sure whether its your source fluid or the upresFluid. The second demo vid seems to hide that a bit better because its blowing out the detail but the first vid looks 20% larger in effect because of the detail.

Then again it can be hard to tell whats going on with all that compression loss from upload.

Have you got any hardware shadows on it? silly question I guess. Also what are the dimensions of your container?

[edit]

120x got it.

Well remmber there is one method where you can obtain one simulation but with 2 opacity sliders?! Using SOuP with 2 containers, I know.... why would you want to burden your processors anymore than you have to, but......if you can try and keep a low enough res source container and smoke container then just maybe?


mm dunno maybe overkill...

Quote 0 0
SGIFreak
very nice node and test videos.  can i render this in batch now out?
Quote 0 0
JeremyR
I believe you need to attach a cache file to the upresFluid. MAya wont render a fluid that is disabled and without a cache. So use the cache on your first fluid and hook it up to the upres'er, it wont eveluate but it will render as it should.
Quote 0 0
pshipkov
The safest bet is to cache the upresed fluid.
You can try that hack suggested by aikiman, but some people reported that it is not working for them.

Quote 0 0

Add a Website Forum to your website.