PDA

View Full Version : Gardner clouds HyperVoxel tests



prometheus
07-16-2012, 02:02 AM
Naehh, this summer vacation is raining away, heck the bad weather even got in to Lightwave:)

Checking some gardner procedurals for clouds(dpont rman collection) with hypervoxels again.
Got som artifact clipping at some places I need to fix, and Im not sure how

theres is around 9 nulls with voxels set up in a cluster.

I am using volumetric shadows, but not textured shadows, generally
that gives that extra touch for clouds, but using it for several nulls is a huge render clogger.

Not sure how it looks on a good monitor screen, this is done on my laptop, render just a direct draft from VPR.

no post processing.

all the sunsky plugins from DPont for the backdrop.

not sure if I dare to start final render:)

http://forums.newtek.com/attachment.php?attachmentid=105557&stc=1&d=1342425737

Michael

CaptainMarlowe
07-17-2012, 11:58 PM
I wish I were able to get this kind of result with HV. Could you post a pic of your settings here, or a sample scene ? That would be greatly appreciated.

prometheus
07-18-2012, 04:58 AM
Ill post a pic of some of the settings later today.

Michael

prometheus
07-18-2012, 07:07 AM
heres almost all settings covered, from one single null, there are 8 of them positioned together to form the whole cloud.

theres a gradient in the luminosity channel, but for these, it isnīt activated.

a lot has to do with the lighting angle, and I also change values in move and rotation channels to find the texture shaping as I want, sometimes they do not look as good depending on rotation or position of the texture, so a fly around of one voxel set would probably show that it looks unnatural at some angles.

Im using DPonts sunsky for backdrop and sunsky light, changing turbidity and other values in sunsky affects the lighting,colors etc on voxel cloud too.

use whatever particle size you want to match your scene, but here it is quite small, the particle size you choose doesnīt affect the general cloud look thou.
Edit..the gain, and offset channel in gardner clouds are very sensitive and controls how to cut off the shape, by default the falloff setting in the main voxel tab seems to
do nothing really, unless using nodes, offset and gain,bias and clamping is another area Newtek could work on to improve falloff.


I wish for HV blend with proper tension blending some day, something like the dynamite plugin had, and would be great to have polyons in the same way.
hypervoxels working on full geometry would be great too, this is something modo has implemented on their first voxel system, and I think Ive been nagging for years how I want that in lightwave.

Try and tweak and set the settings accordingly, then you would have to try rotation and move settings by your own eyeballing to get it decent.

Edit..the gain, and offset channels in gardner clouds are sensitive to cut of the texture in order to avoid spherical cut offf, this is something that should be improved within
Perhaps newtek can improve that too someday.
the main falloff channels wich seems to do nothing in lightwave, except if you are using nodes, then it works a little bit.


Good Luck

Michael

prometheus
07-18-2012, 07:14 AM
I reminder,turn off raytraced shadows if you can, and render out clouds seperatly if possible, it will shoot sky high with raytrace on in final render.

also when working with hypervoxels in VPR, turn of volume shadows for faster previews.

Michael

CaptainMarlowe
07-18-2012, 09:38 AM
Many thanks.

Netvudu
07-19-2012, 10:20 AM
I reminder,turn off raytraced shadows if you can, and render out clouds seperatly if possible, it will shoot sky high with raytrace on in final render.

also when working with hypervoxels in VPR, turn of volume shadows for faster previews.

Michael

which is the reason for some render engines like Houdiniīs Mantra using depth map shadows, which are nothing but shadow maps that respect opacity/transparency of the surfaces.

People think that shadow maps are a thing of the past, but for heavy volumetric stuff, point clouds, millions of particles and many fur systems they are the go-to solution if you want to cut render times on a 2000% basis.

We could really use depth map shadows in Lightwave for this kind of stuff. Specially if LW wants to be active in heavy VFX scenarios.
For things like 20 million particles renders itīs the only sane way to go.

EDIT: By the way, very nice voxel clouds, Michael

Cageman
07-19-2012, 10:23 AM
Nice clouds! :thumbsup:

This is the type of stuff that I think you should submit to NT for inclusion in future Downloadable content!

prometheus
07-19-2012, 10:28 AM
which is the reason for some render engines like Houdiniīs Mantra using depth map shadows, which are nothing but shadow maps that respect opacity/transparency of the surfaces.

People think that shadow maps are a thing of the past, but for heavy volumetric stuff, point clouds, millions of particles and many fur systems they are the go-to solution if you want to cut render times on a 2000% basis.

We could really use depth map shadows in Lightwave for this kind of stuff. Specially if LW wants to be active in heavy VFX scenarios.
For things like 20 million particles renders itīs the only sane way to go.

EDIT: By the way, very nice voxel clouds, Michael


Absolutly, agree on that.

One of the + sides, at least compared to afterburner, It is native and no extra plugin, but of course that doesnīt hold up for long, modo just recently got itīs own voxel system, and some developed to work with painting and pushing point clusters directly in scene, and also to work with geometry shapes and not only vertices or particles. I do not know if they have shadow maps in there?

If they could study and steel from dynamite a few things, it would also be great, fireshader, polyons, opengl presentation etc.

when Im working on several null clusters, I would want to select all or some of the clusters and Edit channels at the same time, that would be nice to, instead of copy and paste etc, because that would destroy some of the voxels unique appareance in other channels.

Personally I think, getting these things in there would be somewhat a priority..

1. shadow maps (faster rendering)
2. proper tension blending between particles. ( better overall shape blending within particle and point clusters)
3. Geometric volumetric ( better cloud shapes)
4. distance between particle gradient in all channels possible ( for fading out smoke and better liquid look where particles break off)
5. gradient preset loading and saving. (I like different colors sometimes without destroying hypertexture shapes)
5. fireshader ( takes quite some time to set up hypervoxels and have that realistic)
6. polyons

And thereīs more...not mentioned here.

prometheus
07-19-2012, 10:29 AM
Nice clouds! :thumbsup:

This is the type of stuff that I think you should submit to NT for inclusion in future Downloadable content!

I might, if Newtek shows some more progress on the hypervoxel front, favour and return of favour:)
I would work a little more on them thou, takes quite some time to tweak right.


Michael

Eagle66
07-21-2012, 06:21 AM
not sure if I dare to start final render:)


That would be interesting. I have problems to get HV photorealism looking - and don't know if there are better settings in LW or you have to fix this in any kind in post?

HV photorealism for VFX integration is prove elusive...

prometheus
07-21-2012, 07:55 AM
It will still be a long way from photorealism, If I had gardner clouds or lightwave procedurals in Vue, that could look really good, Im not fond of the metaclouds in there for building up dedicated clouds thou.

So except for some adjustment improvements in Hypervoxels itself, we reallly need an atmospheric & Lighting system to go with it, and preferably a cloud shader that behaves realistic too, as vue almost do, however.. I would still like to have the shading options that are there today since that gives more options than if I would only have the same shading as in metaclouds for vue.

working on some particle stuff now, so I will take a look at building clouds again later.

I think A method of simulating with turbulenceFD for clouds would be the most realistic method, and faster.

Michael