WIPs: CTZn

[CTZn]

Holy moly : O I want that for C4D!

all your bases are belong to us ! Surrender your skills !

[OnoSendai]

Well, since you like it here's frame 60 !

http://www.worldparallelftp.com/adam/me ... HR.60.pigs (9.56MB)

You may want to change or remove the spp limitation.

The simulation is a tweaked flame preset in Maya. Dealing with emission temperature may be more complex to implement for me, I'd take any hint, C&C's.

Nice. It renders pretty fast.

[StompinTom]

the video;

http://www.worldparallelftp.com/adam/me ... 40x480.mp4 (4.4MB)

I think the exporter is practicable for some really HD fluid simulation... in a still image

For the record I once waited 28h to export one frame (granted, something went wrong and I worked the code out since).

Added voxel space jitter and enough UI controls to tweak the final look. On shading, there's room for some ISL based on distance to the fluid emitter. that's plain diffuse_transmitter with no medium.
I should clamp the spheres radius again; though they add drama, big spheres are also reducing the shading accuracy.

Looking forward for fun ways to use point data in Indigo.

Looking sweet. How many spheres at any given time?

I've got datasets that are 2B+ points in size... Blendigo takes ages exporting even a couple of million particles. I imagine there would have to be some other way of handling larger sets?

[OnoSendai]

Hi CTZn,
You could try using a specular material, IOR 1, with some scattering and/or absorption.

This render was absorption coeff = 1m^-1, same for scattering coeff.

It renders a lot slower though

[CTZn]

I did in earlier tests, and will further. It's good for thick smoke !

Because I also used a strong forward scattering (g spectrum of 0.85, clouds approximation) I was having an issue with fireflies. MLT was hardly helping in regard with my rig power. Looking forward for its announced enhancement ! I've been playing with ROND to avoid TIR on nearly parallel intersections since.

I'm still in an evaluation/dev process so I may be emphazing on previz for a little while (DT was fast) ! I wanted to try a DT/null mix based on ISL noise. Bump mapping does some trick too.

By the way Ono, should the IOR be 1.0 or 1.0029 ?

[OnoSendai]

You want to use IOR 1.0 exactly. it's very important for fast rendering.

Here's another one, with scattering and absorption coeff 0.1:

[CTZn]

You want to use IOR 1.0 exactly. it's very important for fast rendering.

Here's another one, with scattering and absorption coeff 0.1:

Hence my concern with IOR, thanks for clarifying !

Subsurface scattering is definitely the way to go. I would expect clamping spheres to a diameter of 1.4x the voxel width (w/ jitter removal ?) to homogenize the medium setup somewhat, allowing for step size/ROND optimizations.

Though that should only be a concern with very large datasets I'd assume. Will resume works on this later, there is some more available data worth exploring !

[Zom-B]

Here's another one, with scattering and absorption coeff 0.1:

I let this one cook a little longer, used PT with new sky model and different tonemapping:

[lycium]

Wow, this works really damn well and could be used on a much bigger scale! Had an idea for an algorithm to distribute these "volume spheres" in a big big grid of data, so it's more adaptively sampled... maybe a stepping stone to doing true hierarchically refined volume grids

[CTZn]

Here's another one, with scattering and absorption coeff 0.1:

I let this one cook a little longer, used PT with new sky model and different tonemapping:

Point made, thanks Zom-B !

lycium, octrees you said ? Hehe, why should I say no* !? Where the displaceable spheres already XD (woops) ?

edit: ah, that would rather be from point datasets (density/homogeneity to octree resolution) ?

* Because Maya does not manage octrees natively !

[Pibuz]

..I wonder how could I possibly add such a feature in archviz....

AND DON'T SAY THE CHIMNEY'S SMOKE!

[CTZn]

godzilla visiting the house with his relatives ? Not sure that "a volcano" was a better answer



more seriously, can spheres do motion blur ? Either what I may pipe instancing in. There's sub-frame velocity data around, not that I'm sure to succeed at this but that would be awesome !

[CTZn]

godzilla visiting the house with his relatives ? Not sure that "a volcano" was a better answer

For the matter, I may give a volcano scene a try. Doing research on getting more details eventually, nothing to show yet. This might even digress into further development.

[CTZn]

Hi, I have exported frame 59 from a new simulation run at 3x the previous resolution; however, the sole IGS weights 860MB (nearly 6 millions spheres !) and my pc does quite not have enough RAM to test-render the frame, it's swapping disks to death.

So if you have 12 or more GB available (my rig's got 8GB) and would like to render that frame, I can link you to the zipped version (141MB).

Please PM me for the matter.

for the record, the 120 IGS's for the anim I posted weight 10GB together

[Zom-B]

Here we go:

Loading and preparing scene took 72sec ^^

It seems you reduced the spere size a lot, and by that the smoke is way thiner as before.
You also can notice some "line artifacts" that seem to happen because of the simple box grid arangement of the speheres, so that gaps can be viewable...