Render-time Resolution Scaling of Smoke/Volumetric Effects

Sometime back we finished a show where the smoke was not looking as detailed as we would have liked. I then started working on an asset which is an attempt towards resolving this issue.

This asset is able to work on low resolution simulations modifying/adding more detail on it.
It uses the velocity of the fluid to advect the fluid (density field) and add some displacements (along a divergence free vector field calculated at render time) all at render time and is virtually resolution independent as the higher resolution volume is procedurally created from the lower resolution volumes.

For the following tests, the only fields cached out were:
1) Density
2) Velocity
3) Temperature

This approach sort of bypasses the need to resimulate at higher resolutions and
can be divided on the farm much like any ordinary rendering jobs (since the method does not require any information from/about the previous/next frame )


Rendertime UpResing of Volumes, comparison. from bhavesh pandey on Vimeo.

Will do a few other tests as well and update later.

Fire and Water interaction ( Evaporation, Condensation etc )

This is one project I did some time back before running off to a splendid holiday with family.

These are few tests where Fire and Water simulations interact with each
other simultaneously .i.e both these elements are simulated together at the same time
in a single setup (which uses a bunch of microsolvers to move/communicate/manipulate the data around).

In the following video we have fire heating up the water and churning it up as well
as water evaporating at the boiling point.


Boiling Water from bhavesh pandey on Vimeo.

In the video below we have, fire inside a fictitious cold container (sounds ridiculous, I know)
which causes condensation as well as evaporation of water which gets heated beyond the boiling point (as can be seen in the form of smoke trailing behind the heated water).





This is quite interesting, since the usual approach is to simulate these elements separately and then stitch them together after the fact.
I have a few other tests and would be updating on this soon (hopefully....)

Cheers....

some hush hush work..Masters project.

I feel as if i have been reborn..last few weeks were really busy and exhausting. But they have been quite productive nonetheless.
I have been experimenting to my heart's content and the results are looking good.
Following are some of the images from each of the shots I worked on (just 3 of them.)





I will be posting the link for the video (and the "how to" details etc) in some time..
need to touch up somethings..should be done in few more days :)

Elementary....

I have been working on Naiad for some time now and testing few things on the go.
As of now, most of the tests have similar principles as I need to check all the possibilities for one of my projects but as soon as I can get over with this, I guess I have all the time to work on some better/interesting effects.

This is a test where a water character is running on a pavement and sort of disappears from the scene towards the end. At the moment I have just stopped the emission at a particular frame (as an alternative) but I guess I could make it behave in a different way for that.


I guess I'm slacking off..the idea of writing another post doesn't really excite me for now(lots of work left)..so thought would probably add in another naiad test here..

Naiad..Meshed and Rendered ..

In this setup the body is levitating in air mimicking the movements of an animated torus and spills only if the velocity is greater than a threshold. Simulation done in Naiad.

Here's a test with a meshed out render.
Rendered out with Houdini's Mantra renderer.


The Catmull-Rom filter gave me best results for this. I have started to prefer it over the gaussian 2x2 which slightly smoothed out/blurs my renders a bit.

Naiad..another jab at animation driven fluid dynamics.

This has been quite a busy week. I'm working on a project as well as trying to get a grip on Naiad and its workflow. Its been great so far. Somehow it has made the under the hood operations of fluid simulations even clearer to me. I guess trying to work on an unfamiliar software and interface sort of coaxes my mind to reason out with every step that I take. Forcing me to think why should I go with a particular operator and why not to go with some other operator.

Anyways, I was always stuck at a point where I did have the fluid imitating the movement of the torus but it was more like plastic deforming after some time. There were no fluid movements visible.
With help from Michael Davis at the exotic matter forum I was able to fix that.
Here's a flipbook of it:



 I will continue to keep testing with it and post some results here.

test with Naiad!

Well, I have been doing tests on naiad to get a hang of naiad workflow and operators.
this is one of my first usable tests (had a few which are not really worth putting up!)
and since there have been plenty of water based storm effects, I figured it would not
really hurt if I added one more in the count :)



I intend to implement the animation driven simulation setup I did in houdini in Naiad
but I haven't really reached there.
So until then, happy simming :)

animation driven fluid dynamics....atlantian toroid!

This is just a part of RnD I was doing on fluids for making them Art Directable while still preserving their fluid nature. Sort of creating a water body which has a consciousness of its own.



A torus was animated and the objective was to try and get the fluids to move the way the torus was moving while still behaving like water/liquid. This animation drives a water/liquid body while its still simulating thus creating a fusion of both the effects to an extent.
(I will try and get this rendered in half HD as soon as I get some more time.)

The whole setup is procedural and implemented in Houdini using FLIP fluids.A custom setup was built to get rid of the recursive nature of the split pop which was then used to process the fluid for splashes.

Now I am trying to implement the same system in Naiad. I am really happy to see its performance and solve quality. So far I have been able to control the movement of the water/liquid body using velocity fields and Field NEL NOP (Naiad Operators) so its sort of suspended in air and moving as it should be but I'm clearly missing something here. Will put up the results as soon as I get some more time.


PS: I really hate it when I cant seemingly preserve the render quality while converting image sequences to movs in Linux. But thats another story..
HD video now uploaded on vimeo :)

chao.. :)

Fluid Simulation....(of a Viscous and thick fluid)

This literally ate my head off....the simulation was pretty straight but what actually took time was retiming entirely and deleting the stray particles..since i needed the mesh to also stretch I lost the luxury of "Exclude the Lone Particles" option :(

this is a low res version..the original simulation turned out to be very fast but the movement of the fluid was pretty good....so I retimed the entire simulation using a custom Retime node (scripted in Python by Michael Rice) further there were quite a few stray particles which were deleted using ID attributes using a customised setup within Houdini. This could also be done using one cheap solution but its very heavy on the processing side..(sounds so Ironic....haha).

Also, I'm working on few other effects..will post updates once they are finished.

few of the experiments with the color and look can be seen in the pictures..please note that these were low resolution test render pictures

The Biblical Flood..of CHOCOLATE :D (fluid simulation)

hhmm..things moving pretty good and fast this once..(is it because my college is over??..hmm*thinking* hehehe)

orrighty..so here's the Fluid Simulation I had posted last time....just few changes though

I made it lot more Viscous..I wanted to give different look kinda like a chocolate..

I am facing some issues though..as is always the case..things tend to get nasty with me....good thing is..i'm nastier than them hahaha

Ok..lets move on..

I have some Self Shadowing artifacts as you can see in the picture. Issue is if I increase the bias and raytracing intersection distance..I'm losing a lot of shadow detail..

Wanted to see both the pics (i.e. with and without Motion Blur) just for fun of it..

I have thought of some camera movements and stuff but nothing specific as yet.

PS: well thats enough about 'the Biblical Flood of Chocolate' :D

The wetmap implementation is still not complete..but its almost there..(trust me guys I know it..I have done it before :P )

(yeah i know..still not there with the ART of cracking FUnnY jOkES )

flood in the bathtub??

i'm getting terribly restless..reason??

one of the most frustrating thing for me is no longer there to eat my head off.

Since I had worked on Realflow (which I really liked), I could not really get my grasp on the particle fluids in Houdini.

I cant really blame Houdini for this....its all my misunderstanding. I was trying to work on Houdini the way I used to in Realflow. Having said that I must say, Houdini and Realflow are very much alike if some of the things are cleared when you switch. In a way Houdini gives far more control to the user as compared to Realflow.

I'm not taking anything away from Realflow because I simply cant. Its made for fluids and dynamics....I do also feel its faster than the fluids in Houdini.

I use Houdini fluids for the simple reason: MORE CONTROL.

Realflow did not have a official Houdini export until now (there's a beta version Realflow plug-in for the same..for more details on this please cheack Odforce.)

And since I have been working on Houdini primarily for everythin, I cant really use RF fluids.

Also there's a lot of control which if leveraged properly, can get you amazing results.

Anyways, I used to be really irritated because of lousy and stupid looking meshing of the Particle Fluids....which I again confess was my lack of understanding.

Finally while working on this personal project the solution dawns upon me.

Its got to do with the Particle Seperation.

If you have a high seperation (which I used to keep high to get better speed) the partcles will be farther away from each other and thus when meshed, the mesh turns out to be entirely stupid. (come on I could draw up a better mesh on paper than THAT).

This once, I kept the separation lower, keeping sure it not unnecessarily slowing my system (a value lower than 0.1 but higher than 0.01 usually gives good performance/quality ratio) I tried to get the desired look changing parameters like Bounce, Friction etc.

Once I got something similar to what I wanted, then I would try to up the resolution.

Well to be honest, I also cheated a bit as far as resolution of the sim is concerned.

Ok, its still in a Work In Progress image so I have got a screen capture and a quick test render done (which is nothing fancy: no motion blur..no texture etc).

almost done with the wetmaps implementation. (Houdini doesnot generate wetmaps straight out of the box..but to learn how it could be done is even more fascinating!)

Will update later :)

PS: please ignore the title....it's another side effect of my self imposed isolation haha :D