Metacloud Raymarcher

[AndrewVK]

Here my test of pure reymarcher through metafield.

As for me, it works better than i3d in "rocket trail" like situations,

because no need to worry about i3d resolution.

Since rocket trail is just sequence of bgeo saved to disk, i have a one BIG problem...

I have no idea how to obtain Motion Blur.

Any suggestions on how to fix this are welcomed.

I`m afraid it is not possible at all, but who knows... :coffee1:

MetaCloud.zip

[sibarrick]

Nice render What are the render times like?

Sorry can't answer your question, but good luck.

[Ratman]

Don't know about the motion blur stuff, but ohh man is that an awesome shader , just giving ya a complement.

[AndrewVK]

Thanks guys

AMD 3700+

2Gb RAM

Step Size: 0.05

Self Shadow Step Size: 0.1

Self Shadow Noise: On

Render Time: 8 min 35 sec...this is the price of "no need to worry about..."

On the other hand the time to prepare 1024x1024x1024 i3d is much longer.

There is nothing special about the shader...it uses almost the same "engine"

as VEX 3D Texture Fog.

[lynbo]

That is really good Andrew.

Motion Blur is going to be a pain with an object from disk.

No velocity vectors.

You might try using the metaballs in a seperate render that gives you 2d motion vectors and do a post process motion blur

using reel samrt or smooth kit plugins.

My 2 Cents

[lynbo]

Ooops that should read

"reel smart" not samrt

the link is

http://www.revisionfx.com/mblur.htm

http://www.revisionfx.com/smok.htm

[edward]

Can't one use houdini to generate motion vectors too?

[renderfox]

Adrew,

at www.renderman.org you could find paper, described pretty fast technics for motion blur of volumetrics.

Somewhere in SRT pages (sorry, I don't remember where exactly)

[Rafal123]

Hey Andrew cool render. How did U generated metacloud.bgeo ?

[hoknamahn]

Hey guys. It is possible to write out velocities of metaballs but here is one problem - which velocity must be rendered? Or velocities must be integrated? In last case I see other problem - if one "smoke layer" have velocity = 1 and other "smoke layer" have velocity = -1 we can get result velocity = 0.

[AndrewVK]

It is metabals copied over points (particles) (non renderadble "model" object)

Ok...i will try this:

export velocity vectors to metaballs

Import vector when density registered for the first time....

output this vector to deep raster...blur in comp

[3__]

hey these render-times are approaching afterburner...

Is it possible to force metaweight to use the source

sop rather than a file?

also what's causing the hard edge when I reduce density?

-cpb

[AndrewVK]

Is it possible to force metaweight to use the source

sop rather than a file?

Unfortunately no...

op: syntax doesnt work

also what's causing the hard edge when I reduce density?

Can You show picture?

My guess...unhidden original metabals.

[3__]

so files are used because it *couldn't* work

with the sop? just curious...

also here's the pic. changed settings are:

density: 0.8, turbulence: 4. no self shadows.

removed masks to all but 1 light placed behind the cloud.

if you just reduce density, turbulence and switch

off self shadows on the posted file you'll get a similar

result. looks like clamping or an additive effect.

-cpb

[AndrewVK]

so files are used because it *couldn't* work

with the sop? just curious...

Exactly...

Is red line drawn by You to show edge or this is mantra?

I can not reproduce red line...

Hardness of the "edge" depends on how density

is distributed inside the volume.

Perhaps there is to much metabals...so the density is MUCH greater than 1

I`m think You will get the same result if You create i3d from this metacloud

and render it with VEX 3D Texture Fog shader

[peship]

there is a very old technique to get moBlur without actually render it.

just render subframes, this will produce a lot more images,

then timestretch them in compositing - the interpolation will give you the motion blurred look.

timewise the difference is not that big - rendering volumetrics with motion blur usually takes a lot of time.

rendering of non-motion blurred volumetrics is faster, but rendering bunch of additional subframes will rise the render times.

so ... it's so - so.

[AndrewVK]

I have reworked shader slightly

Now it uses metamarch function, perpoint colors,life depended opacity and noise amplitude.

However i am still fight with velocity.

How to get attribute of single mataball from a large amount of metaballs?

Suppose i want to get velocity...

int found = 0; // Velocity Not Found yet
...
...
...
while (metamarch(index, fname, p0, p1, bias)) //Lets roll
{
 ...
 ...
 ...
 for (i = 0; i < nsteps; i++) // Something found...lets test it
   {
	...
	...
	...
	opac = density * clamp(dfunction(fname,p0+noise,dexp),0,1); // Is there some density?
	if (!found && opac > 0.0001){ // If not found and there is some density then...
	  v = getvelocity(fname,"v",p0+noise); //get velocity
	  found = 1; // I found it 
	}

When i call getvelocity i am sure that "p0+noise" inside one of metaballs

So...theoretically i can get it by calling:

handle = metastart(fname, p0+noise);

metaimport(handle, aname, p0+noise, result)

However i got nothing result is always zero.

I am forced to use such function to get something

vector
getvelocity(string fname, aname; vector p)
{
  int	 handle;
  vector  result = 0;

  handle = metastart(fname, p);
  while (metanext(handle))
  {
	if (metaimport(handle, aname, p, result))
		break;
  }
  return ow_vspace(result);
}

Whats wrong??? Can anybody enlighten me?

DivX.zip

[MADjestic]

Hmm...

It would've been nice if fog context supported some surface context functions, like 'forpoints()' - so that one could access surrounding points and query their attributes. This way getting velcity attribute would be piece of cake.

Good luck in your perseverance.

[MADjestic]

On the second thought...

I've just tried declaring pcopen(); function in VEX Fog context and didn't get any "function declared outside of context" messages - so I thought it was possible to use this functions to query nearby particles with which the metaball object wascreated with? Then you could assign V attribute of a given P in your Metmarch shader by copying it from nerby particles... It's just in theory - can this be done?

I would try it myself, but I am inexperienced with pointclouds. Still - if anybody who has more knowledge sais that it is possible - I'll try this as well. So, wat do you think, guys? Andrew?

[edward]

A fog shader (aka Atmosphere Shader only has access to the current pixel colour I think. So at this (last) stage in the shading pipeline, I don't think it makes sense at all to try to access your point position as you're not shading geometry anymore.