Hello all,

I'm in the process of setting up a workflow to optimize an ink-in-water type rendering for a video clip I have to produce.

The technique I'm using for the moment is quite classical, and consists in advecting particles to a smoke simulation, and then multiple cache by wedging and randomizing the seed of the particle sim. Depending on the situation, I am around 10/50 millions of particles, and even with that quite huge amount of particles, even with the motion blur activated (I use Redshift), I'm approaching the rendering I am looking for, but not quite. The final render is still a litlle bit grainy, and I don't have perfectly the same whispy effect as real ink in water.

So my question is, what would be the ideal workflow to achieve this result ? All the ressources available online are just scratching the surface on this question, and I would need something a little bit more advanced... I've seen some Krakatoa rendering that reach my needs (like in the famous CCTV ink ad, where the initial ink drop effect is absolutely gorgeous, that's what I'm looking after), but I don't konw exactly how to replicate the workflow. I can assume it has a smart approach of partinniong and caching, but I don't know exactly how.

One of my theoretical tracks would be to generate a first advection of particles, and to find a way to make the complementary partitions, whatever the number, compute the intermediate position between a particle and the previous one in the first simulation, and to position each particle of the sim+1 at this intermediate position, and to be able to multiply this operation
But I don't know exactly what the workflow would be for that. If that sounds like a good idea, how would you go about it?

Because by using wedging to randomize the seed of each simulation, the position of each particle in sim+1 is completely random...

If you have any clues for all this, I'm very interested! Or maybe even Houdini is not the ideal solution, feel free to point me to a software that can do the job...

Thx in advance!

Hey frBorges34,

I feel your struggle. I worked on the CCTV ink spot back in the day. We used Fumefx and Krakatoa, and yes point count was ridiculously high. I think Krakatoa is still better suited for the task than Houdini is. And it's free now? (I don't know it exactly)
It has all the functionality built in to cache millions and millions of points and render them.

Cheers
CYTE

Hi Cyte,

I'm very happy that I can talk to you, the CCTV ink is an absolute work of art, I can't express how much I admire the work done ont this video, the motion design, the visual ideas ! I looked at it in detail, at all possible speeds, an alarming number of times. I was so impacted by it that the first scene of the clip I made is a direct tribute to your work.

If I'm not mistaken it is indeed a mixture of fumefx and Krakatoa, but the rendering also mixes both particles and smoke, right? For example I have the impression that some scenes are entirely made from advected particles (the swan flying away, the kungfu scene etc), while others are pure smoke (like the carp in the beginning of the clip), or am I wrong?

I can't understand what is the current model of Krakatoa, I saw that the aws Thinkbox products were free, but I don't understand to what extent, nor if it is still only linked to 3dsMax (which for the moment I don't own, I have at my disposal C4d and Houdini currently). So I don't know yet how to proceed...

If I may ask, What were you dedicated to on this clip ? I have so many questions for you haha

That's kind to hear, yes it was a nice project.
I was working on the whole system, as I was the only one experienced in 3DSMAX and my Collueges worked with C4D.
It was all rendered in Krakatoa, and Fumefx was used for the simulation. At some points, real ink was added in compositing.
I think Krakatoa is(or at least was) also available for C4D, so you could give it a try.

Cheers
CYTE

Okay, it appears it's not available for the last versions of c4d,t the last supported one being the S26 version.
By any chance, do you have any idea how does Krakatoa is partioning, caching and rendering, under the hood ? Just to estimate how dificult it could be to mimic it in Houdini... Or maybe some documentation ?

Sorry man, I'm an artist I have no idea when it comes to this stuff. I guess you can mimic the caching part with TOPS quite well.
For rendering Krakatoa is neat because you don't have to load all the caches anywhere. Krakatoa loads them only when you render them. Plus you get the desired look with the points by adding them with super small alpha values like 0.00001 and then the wispiness builds up. for that in redshift, you need a very high amount of Transparency steps. that slows down the render time. I think it's all possible with redshift and Houdini but with Krakatoa, it renders quite fast and effortless. Good luck.

A Krakatoa integration for Houdini would be a match made in heaven. Maybe yet it's open source somebody can make that. There is no alternative to it when it comes to Rendering Billions of points as far as I know.

If I had to do it nowadays I think I would render the volumes directly. I made a test out of curiosity. take a look.

Yes, actually, that's what I've done for some pre-shot testings, and it surely would cut the whole time (and hard drive) consumming process by half. My heart is still balancing between the two solutions, maybe I'll make a mix of both. I found the advected particles a little bit more realistic, but maybe it's just a matter of setting the RS volume shader the right way - way that I didn't perfectly found for now !

With a high count particles nb, motion blur activated and properly set, and optix denoiser activated (using what is usually its flaw: its propensity to smooth out the pixels, as an advantage here), the results are quite convincing. It's just it take so much sim time and HDD space that it's a real struggle, and not so simple to art direct...

I would do a medium res pyro sim, do your advected particle wedges, then rasterize those wedges into volumes.
You'll get the nice tight strands you want, without needing 1billion points, it won't be grainy either.

L

lewis_T
I would do a medium res pyro sim, do your advected particle wedges, then rasterize those wedges into volumes.
You'll get the nice tight strands you want, without needing 1billion points, it won't be grainy either.

L

I've just tried that, but I can't manage to have something more convincing than what I have with a simple smoke, but maybe I'm missing something here.

CYTE
If I had to do it nowadays I think I would render the volumes directly. I made a test out of curiosity. take a look.

By the way, can I ask you a little more about the config of this pyro sim ? Do you use the pyro sop or the classic way ?

I used Axiom. It's the default Parameters except Bouancy goes down.

Cheers
CYTE

CYTE
I used Axiom. It's the default Parameters except Bouancy goes down.

Cheers
CYTE

Thx !