PDA

View Full Version : Volume Hypervoxel shape and Node Volume



serge
10-27-2008, 02:10 PM
I've been experimenting with Dpont's Node volume to see in what ways we could change the shape of a volume hypervoxel.


First I tried Dpon't boolean node on the opacity of the texture, but that doesn't seem to work (hey, you never know). Second I tried Spot-Info. If we could use for example the "Bounces" or "Incoming Eta" outputs, we could cut away parts of the hypervoxel with another (transparent/refractive) object, and thus create any shape we want. That didn't seem to work either, so my quick conclusion was: Hypervoxels and Spot-Info don't work together because it's a volumetric and not a 'normal' surface object.
However, further experimentation revealed that this is not true. Some Spot-Info outputs actually do work (see examples below). "Object/World Spot" and "Ray Length" for example can be used. Which leaves the question: why do some Spot-Info outputs give the info you expect, but others not? ("Normal" or "Polygon Side" are understandable, they don't make sense when dealing with a Hypervoxel, but "Bounces" or "Incoming Eta" should, no?)


See attached images:

Image 1:
Cutaway (texture opacity) using World Spot X/Y values.

Image 2:
Cutaway (texture opacity) using World Spot based on distance to nearest point from reference object (Dpont's Nearest point node). Before you say: "hey that's cool and useful", you should know that this setup can affect rendertime drastically. Although you're cutting away part of the hypervoxel, and might expect a faster render, the extra calculations can increase rendertime enormously (depending on how many points the reference object has).

wildr3d
10-28-2008, 12:13 PM
I like the first image, if I never need to make a dust bunny version of pac-man, this will work!

I hope Denis finishes the NodeVolume plugin, I think
many would find them useful.

dpont
10-28-2008, 01:10 PM
nodeVolume and nodeTexture are Texture plugins
so they couldn't have access to shader infos,
there are limited to local and world spot
position and can't get spot normal,
Boolean node could not work
without normal infos and need geometry.

Raytracing is theorically possible with HyperVoxels,
but through shader plugins.

Denis.

serge
10-28-2008, 02:50 PM
...a dust bunny version of pac-man...
Yes, that too :)! But I was thinking more of situations like smoke hitting a window, or a droplet of ink in a glass of water. In these cases we need to cut off the shape of the hypervoxels to prevent them from penetrating the object. Having Spot-Info "Bounces" or "Incoming Eta" work with hypervoxels would mean we have a good solution for those cases.


...Raytracing is theoretically possible with HyperVoxels,
but through shader plugins.

Denis.
Does this mean it's possible what I want, but then a Node Volume Shader Editor has to be written?

dpont
10-28-2008, 03:27 PM
Not for Hypertextures, but for texturing with HV Shader,
for sclupting volume itself I think that it is possible
but in a new/different volumetric plugin with a ray-marcher.

Denis.

Edit:...or an enhanced HV version 4.0?

virtualcomposer
10-28-2008, 03:44 PM
It looks like a toilet paper roll that has been left out in the rain and then chewed up by the dog. LOL

zapper1998
10-28-2008, 04:27 PM
Could you share your scene, Mr. Serge..

Please ??

Michael

serge
10-28-2008, 05:42 PM
Not for Hypertextures, but for texturing with HV Shader,
for sclupting volume itself I think that it is possible
but in a new/different volumetric plugin with a ray-marcher.

Denis.

Edit:...or an enhanced HV version 4.0?
Thanks Denis. Yes, HV-4.0 would be very welcome indeed :).


It looks like a toilet paper roll that has been left out in the rain and then chewed up by the dog. LOL
Uhm, yeah I guess so. :D


Could you share your scene, Mr. Serge..

Please ??

Michael
No problem. Scenefiles attached. You need Dpont's Node Volume installed, plus DP-kit (for Nearest Point node)!

I've combined the two nodal setups (for the two images above) in one node-tree. To get image-1 leave the nodes like they are. To get image-2 plug the Logic(3) node into Opacity. (The Node Volume editor you find under HyperTexture.)

Panikos
10-29-2008, 12:18 AM
Follow this thread to find out how to enclose Volumetrics in a poly-shape

http://www.newtek.com/forums/showthread.php?t=29665&highlight=Hypervoxels

serge
10-29-2008, 09:17 AM
Panikos, thanks! I must have missed that thread in my search. You must have missed a few recent threads where users asked for similar effects ;).

Seems like a good workaround (although it renders real slow). I don't understand why this works though. "Unseen by camera" to me means "unseen by camera" whether it's inside a transparent object or not. Is this a bug? If it isn't, could you explain why it works?

serge
10-29-2008, 09:21 AM
Off topic: this post is just to get rid of my "666" post count.

Panikos
10-29-2008, 10:43 AM
Hi

Unseen by camera, if applied on the HV_Object, makes this particular volumetric invisible to the camera but visible in reflections/refractions.
That is, its there but the camera cant see it.

So, once you have your volumetric tuned, you hide it and then you define the shape/housing of this volumetric.

Its a well hidden feature that nobody commented in manuals on anywhere, I was lucky to chase it, catch it and share it :)

Volumetrics in LW are calculated using Ray marching, i.e the ray travels inside boundaries. In Volumetric Lights is a predefined shape, in Hypervoxels its a sphere. I reached to these conclusions when I was experimenting with LW7.* and monitored the various LW Buffers, (specifically the Z-Buffer)
During that time HVoxels didnt work with DOF and they were not present in Z-Buffer. Also the Z-Buffer included only Polygons.

I asked for Newtek to open the Z-Buffer for access to plugins like Sasquatch and I also asked to have Volumetrics in Z-Buffer. Fortunately this happened :)

Panikos
10-29-2008, 03:57 PM
There is a outstanding issue regarding HVoxels and Z-Buffer.
In Z-Buffer HVoxels appear as solid spheres. The actual depth per pixel is not calculated most probably due to speed reasons. The result is not accurate. There is no immediate need for fixing this, or probably HVoxels will be revised later in the future.

serge
10-30-2008, 07:47 AM
... or probably HVoxels will be revised later in the future.
Most certainly, at least Jay Roth has stated that a big overhaul is planned, but that it's going to be a big job. In the meantime we users just have to find ways to work around the limitations. Thanks again for sharing this technique, Panikos. It's certainly helpful in some situations.

prometheus
11-03-2008, 07:15 PM
he..yeah..Ive just learned how to create volumetric objects by implicit surfaces..in Houdini apprentice.

http://www.newtek.com/forums/showthread.php?t=90926&page=4

That would be welcomed for the next release of hypervoxels.
Newtek could start with the polysize and metaball blend for volume mode thou..like the Dynamite plug..and then advance it further.

Michael