Clouds - Simul software technology

146 views
Skip to first unread message

Chris Johnson

unread,
Feb 14, 2011, 11:17:19 AM2/14/11
to soft...@listproc.autodesk.com
Gotta do some clouds...just starting the research phase to see what's
out there.

http://www.simul.co.uk/

Anyone ever used this...last version of support was XSI 7.5?


--
Chris Johnson
3D Supervisor
Topix
www.topixfx.com


Pete Edmunds

unread,
Feb 14, 2011, 11:22:06 AM2/14/11
to soft...@listproc.autodesk.com
Binary alchemy volume and shader plugin springs to mind

http://www.binaryalchemy.de/

Marc-Andre Carbonneau

unread,
Feb 14, 2011, 11:36:26 AM2/14/11
to soft...@listproc.autodesk.com
What kind of clouds are you needing? What's the use of those clouds?
Have you checked Ozone 5? It's the clouds and atmospheric part of Vue by E-on Software.
http://www.e-onsoftware.com/products/ozone/ozone_5.0/

MAC

Chris Johnson

unread,
Feb 14, 2011, 12:07:57 PM2/14/11
to soft...@listproc.autodesk.com
We have one artist working in Vue right now. How do you pronounce that
(view?) .

I'm yet to try the Binary alchemy shaders but I will. We've tried max
and Maya Fluids and our Compositor wasn't happy with the results...so
we're still in a trial stage and looking for options. They have to be
pretty art directed so I don't think a volume shader is gonna give us
the control you want. We'll give it a try though.

Any other suggestions welcome.

tak...@earthlink.net

unread,
Feb 14, 2011, 12:12:36 PM2/14/11
to soft...@listproc.autodesk.com

There's also Terragen, you can export cameras to it as Nuke .chan.

-T

Christopher Tedin

unread,
Feb 14, 2011, 12:26:50 PM2/14/11
to soft...@listproc.autodesk.com
Do you need "fly-thru" or background environment? Either way, you can't
go wrong with Vue. I use Vue 9 Studio to make great environments. The
spectral system is pretty good, and you can animate fly-through
animations. Should be able to bring in an XSI camera and
match-move/composite. Or, get the xStream version and plug it into
Softimage. Works with Mental Ray right inside Softimage. (Although I
haven't tested it, and don't have this version.) Rendering might be a
bit slow.

Chris Johnson

unread,
Feb 14, 2011, 12:45:01 PM2/14/11
to soft...@listproc.autodesk.com
Not fly through but fly over head just above camera. We found with Maya
and Max the clouds had that default volume whispy edge look...we want to
get something I think looking a little thicker and at the same time have
the control of placement.

Thanks for the suggestions.

Marc-Andre Carbonneau

unread,
Feb 14, 2011, 1:07:07 PM2/14/11
to soft...@listproc.autodesk.com
I'd go with basic background clouds in Vue and use Vue's metaclouds to "art direct" specific cloud shapes where needed.

-----Original Message-----
From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Chris Johnson
Sent: 14 f�vrier 2011 12:45
To: soft...@listproc.autodesk.com
Subject: Re: Clouds - Simul software technology

Stefan Kubicek

unread,
Feb 14, 2011, 2:07:12 PM2/14/11
to soft...@listproc.autodesk.com
Since external apps haven't been ruled out and in case you have Max, you might want o look into DreamScape, Afterburn, FumeFX (all http://www.afterworks.com/) or
PhoenixFD (http://www.chaosgroup.com/en/2/phoenix.html)

> Gotta do some clouds...just starting the research phase to see what's
> out there.
>
> http://www.simul.co.uk/
>
> Anyone ever used this...last version of support was XSI 7.5?
>
>
>
>


--
-------------------------------------------
Stefan Kubicek Co-founder
-------------------------------------------
keyvis digital imagery
1050 Vienna Wehrgasse 9 Austria
Phone: +43/699/12614231
--- www.keyvis.at ste...@keyvis.at ---
-- This email and its attachments are
--confidential and for the recipient only--

Steven Caron

unread,
Feb 14, 2011, 2:18:18 PM2/14/11
to soft...@listproc.autodesk.com
the first two maybe, but i didn't have much luck getting fumefx to do anything cloud like. i could get a better result faster with ICE and BA volume. phoenixfd i dont know anything about...

s

Chris Johnson

unread,
Feb 14, 2011, 3:45:46 PM2/14/11
to soft...@listproc.autodesk.com
Yeah we didn't have much luck with Fume FX.

Anthony Martin

unread,
Feb 14, 2011, 4:48:48 PM2/14/11
to soft...@listproc.autodesk.com
Built in particle volume/BA volume might very well get you close. You have the opposite problem to me. Last two jobs I did, the client wanted whispy clouds as opposed to well defined clouds, which the built in particle volume tools do quite well.

The problem I always find though is doing wide shots of lots and lots of clouds streaking overhead. Became very cumbersome and unstable (memory maxing out, looong renders times).

My current thinking is: develop particle/volume shader look. Render out big plates of the cloud. Use that to paint a 2.5d matte painting. Have lovely 3d volume clouds whizzing by in the foreground.

Anthony

Morten Bartholdy

unread,
Feb 15, 2011, 7:54:44 AM2/15/11
to Softimage Userlist
We did 23 shots of aircraft flying in and out of clouds last year for a
feature, and went through exactly the same fase as you are in. Cloudwright
is nowhere near there yet, so forget that. Eventually we did all
cloudscapes in Vue with varying succes I might add, and for direct
interaction with hero clouds we used ICE/Volumetrics. These are difficult
for modeling realistic looking clouds seen from a distance, so Vue had to
provide that and for the foreground hero clouds, it was slow, tedious
work.

Vue is best for specific looks and types of clouds IMHO and the best
results (most realistic) was for the ones where the director allowed us to
use a lot of mist which made for very nice moody cloudscapes with
volumetric light. Be prepared for long rendertimes to get rid of or get
acceptable levels of grain/noise. We used Fusion with Revision FX
De:Noiser http://www.revisionfx.com/products/denoise/ extensively too.

We had limited succes with metaclouds - it is very hard to model a
believable thunderstorm with it. Better for little drifter clouds passing
by.

I find Ozone difficult and unstable so we used a workflow where we
exported animation via Vue Extreme, which is Vue inside Softimage, then
used Vue standalone to get the look right and render. For 1080p we had
rendertimes way beyond an hour per frame where clouds fill the frame, and
fog/mist makes for very noisy renders. ALso try and avoid flying inside or
through Vue clouds - it gets unbelievably slow, and it can be very hard to
see if you are actually inside a maybe 2% dense wisp of cloud at times,
but rendertimes soar when you do.

The upside was we had to get me a very fast machine and extra boxes for
the farm for this job :)

My 2 cents - good luck!


--
Best Regards

Morten Bartholdy
3D/VFX Supervisor

Chris Johnson

unread,
Feb 15, 2011, 9:18:29 AM2/15/11
to soft...@listproc.autodesk.com
Thanks again everyone. Your opinions have saved me a ton of R&D time.

I know what your saying about the meta clouds...I'm already seeing the
limitations in "look".

Crossing my fingers on more RAM or a new box!!!!! ; )


--

Marc-Andre Carbonneau

unread,
Feb 15, 2011, 10:21:45 AM2/15/11
to soft...@listproc.autodesk.com
Here's something that Vue expert Dax Pandhi did. Might be useful to you? Only 9 bucks!
http://www.cornucopia3d.com/purchase.php?item_id=8913

-----Original Message-----
From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Chris Johnson

Schoenberger

unread,
Feb 17, 2011, 3:23:16 PM2/17/11
to soft...@listproc.autodesk.com
Hi

Hmm, the funny thing is that the very first reason the volume shader was
developed was because I needed to create a sky (time-lapse movement).
Some images of the first version 0.5 of the volume shader:
http://www.binaryalchemy.de/develop/shd_vol/img_prod/Weisse_Ameisen01.jpg
http://www.binaryalchemy.de/develop/shd_vol/img_prod/Weisse_Ameisen02.jpg
http://www.binaryalchemy.de/develop/shd_vol/img/maya_Cloud_Sky.jpg


If you want to use the build in volume with an "artistic" shape, then I have
a small collection of different shapes you can create:
http://www.BinaryAlchemy.de/develop/help/2d_3d_comparison.htm


The way I made it is just by plugging a fractal piped into the volume shader
and apply the volume shader to a cube.
The making of:
http://www.binaryalchemy.de/develop/shd_vol/vid/Making_the_Sky.wmv


But as someone in the thread already said, if you want clouds until the
horizon, it could get difficult as the memory usage does increase
exponentially with distance.
As you can see in the first images, the clouds are faded out or clipped by
trees.

One trick I used was to create multiple cubes.
So I could just place them where you see sky.
And if the camera rotates, only the cubes that are seen use memory.
And at last I used different cell settings and details the farther away.

Last but not least:
If you really want to define the shape of the clouds, then the best solution
would be to model the rough shape with polygons.
Then use the BA_mesh_distance shader.
I cannot show you my last project, but there I have made an inner building
with columns (imagine the inner view of a church or throne room)

The attached is not the best example, but this one has a polygonal shape.
But it is a good example for an artistic look.


Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night

|> -----Original Message-----
|> From: softimag...@listproc.autodesk.com
|> [mailto:softimag...@listproc.autodesk.com] On Behalf
|> Of Chris Johnson

Wolke_NIX_A03_StereoCamera_Left.1.jpg

Steven Caron

unread,
Feb 17, 2011, 6:58:23 PM2/17/11
to soft...@listproc.autodesk.com
http://www.binaryalchemy.de/develop/help/2d_3d_comparison.htm

is this page in the softimage documentation? cause that helps a lot in visualizing.

i was wondering how would one get a simple falloff of density from the middle of the particle to the edges? i tried to use a cell scalar but had no luck. i essentially want to control that fallof from center of the particle to the edges and use it to blend in the different fractal scalar and cell scalar shaders.

s

Schoenberger

unread,
Feb 18, 2011, 4:39:43 AM2/18/11
to soft...@listproc.autodesk.com
>is this page in the softimage documentation? cause that helps a lot in visualizing.
I am not sure, they wanted to implement everything...
 
The density of a particle has always a falloff from center to the edge. If you do not have one, then the global density in the volume shader is perhaps very high.
You can only control the falloff itself with the switch linear or cubic falloff.
 
And then apply the fractal/cell into the density shape input of the particle shader.
There are some images in the help above with the cell shader (bottom), anything in it you want to achieve?
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 


From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Steven Caron
Sent: Friday, February 18, 2011 12:58 AM
To: soft...@listproc.autodesk.com
Subject: Re: Clouds - Simul software technology

Adam Seeley

unread,
Feb 18, 2011, 8:46:11 AM2/18/11
to soft...@listproc.autodesk.com
HI,
 
I'm sure I'm missing a basic concept here.
 
I've enveloped a mesh and animated it using a chain. 
When I add an Ice Simulation tree, the envelope animation no longer animates. 
I'm missing something quite obvious I'm sure, but how do I get the mesh to keep deforming using the envelope.
I want a simulated Tree so I can run some Weightmap calculations.
 
To see what I mean, run this to set up a simple enveloped anim
------------------------------------------------------------
CreatePrim "Cylinder", "MeshSurface"
Create2DSkeleton 0, -2, 0, 0, 0, 0, 0, 0, 0, 4
AppendBone "eff", 0, 2, 0
SelectObj "cylinder", , True
ApplyFlexEnv "cylinder;root,bone,bone1,eff", , 2
SaveKey "bone1.kine.local.rotz", 1, 0, , , False
Rotate , 0, 0, 90, siRelative, siLocal, siObj, siXYZ
SaveKey "bone1.kine.local.rotz", 25, 90, , , False
Rotate , 0, 0, -180, siRelative, siLocal, siObj, siXYZ
SaveKey "bone1.kine.local.rotz", 50, -90, , , False
Rotate , 0, 0, 90, siRelative, siLocal, siObj, siXYZ
SaveKey "bone1.kine.local.rotz", 75, 0, , , False
-----------------------------------------------------------
 
Then add a Simulated Ice Tree.
 
 
Thanks,
 
Adam.
 
 
 
 
Adam Seeley
Senior Animator, Commercials, UK

T: +44 (0)20 7565 1000  
E: adam....@primefocusworld.com@primefocusworld.com

www.primefocusworld.com


sigLogo.gif
ldn.gif

Robert Chapman

unread,
Feb 18, 2011, 9:06:35 AM2/18/11
to soft...@listproc.autodesk.com
Hi Adam,

it depends on what you want to do really,  normally we would point cache the animated mesh and then apply the simulated Ice tree to that, but for it to continue to work simply you could drag the envelope operator above into the post-simulation stack, or even go fancy and mute the envelope operator and apply the dual quaternation deformation in Ice..
ldn.gif
sigLogo.gif

Orlando Esponda

unread,
Feb 18, 2011, 9:47:00 AM2/18/11
to soft...@listproc.autodesk.com
...And if your ICE tree depends on the envelope deformation you can also drag the ICE tree to the post simulation stack, above the envelope op

Alan Fregtman

unread,
Feb 18, 2011, 12:07:26 PM2/18/11
to soft...@listproc.autodesk.com
It's a feature not a bug. :p As soon as you have "Simulation" in your stack, even with no ICEtrees to be found, anything below Simulation only evaluates on the first frame.

Like others pointed out you just have to bring up the Envelope op into the Simulation stack, below your ICEtree op. Should be fine. ;)
ldn.gif
sigLogo.gif

Steven Caron

unread,
Feb 18, 2011, 3:32:27 PM2/18/11
to soft...@listproc.autodesk.com
i am trying to keep the particle fully solid for a certain range from inside to outside, then right at the edge shift to using some fractal or cell. i will try and send an illustration...

s

Schoenberger

unread,
Feb 18, 2011, 3:46:27 PM2/18/11
to soft...@listproc.autodesk.com
 
Just increase the contrast in the fractal_scl shader.
If you have a default fractalsum, then set min/max contrast to 0,45 and 0,55
Then set the intensity of the "shape density" in the partilcle shader to 1.
 
Or increase the contrast slider in the volume shader (but I would go with tweaking the particles)
 
 
Hint: What a lot of people forget is to use the shader ball.
Works fine to see the contrast in the fractal.
And to see a sliced particle in the particle_density.
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 

Sent: Friday, February 18, 2011 9:32 PM

Steven Caron

unread,
Feb 18, 2011, 3:50:32 PM2/18/11
to soft...@listproc.autodesk.com
i use the shader ball exentsively but set grid so i can visualize the texture/scalar. i am going to try this right now...

Steven Caron

unread,
Feb 18, 2011, 6:23:00 PM2/18/11
to soft...@listproc.autodesk.com
ok i had no luck... i comped this... i took a fractal scalar, with high contrast. 0.45/.55, 2 distortion, and 'small noise', scale of 1. then i just laid a shader ball render of the particle_renderer with no fractal plugged into density shape. i then set the parameter to 0.75.

i want it to be a density of 1 from the center of the particle to 3/4 of the distance away from the center, then i want to blend in a small noise.

steven

Steven Caron

unread,
Feb 18, 2011, 7:54:37 PM2/18/11
to soft...@listproc.autodesk.com
hmmf... image got blocked...
noisy_edge.jpg

Vincent Fortin

unread,
Feb 19, 2011, 12:09:49 AM2/19/11
to soft...@listproc.autodesk.com
I too look forward to create realistic clouds in Soft. Here's some more examples from Holger's website that are interesting.
A lot of those examples are too noisy to my taste and feel a tad unnatural but I'm sure we've got everything for the task in Softimage. Not Softimage but the clouds in Legend of the Guardians were simply hand-made geometries filled with volume. Clever usage of fractals was mandatory but the little touch that made the hero ones stand out is the simulation added on top. A simple localized curl noise for example can really help breaking the procedural look from the textures.
Seeing how fast the community is responding to ICE, the addition of a voxel framework in it would potentially blow the competition away. I really really hope it will happen one day.
In the meantime, Houdini is ahead in terms of control. Here's some great looking examples:
http://dnabox.com/?p=152
But for those who take to heart having a life fully lived, I agree, Houdini might not be an option.
Vincent
P/S: @Steven, my girlfriend just walked behind and asked "WTF, is that a hole in a bathroom?". No offence, I just has a good laugh and had to share ;-)

Steven Caron

unread,
Feb 19, 2011, 12:41:36 AM2/19/11
to soft...@listproc.autodesk.com
lol that got me chuckling!

i have some clouds which i am quite happy with! but i still want to push it further. the houdini examples illustrate scale very well and i feel is the biggest draw back to the volume shader. on top of the large cellular cloud shapes we need noise that is 'displacing' the volume at a much smaller level. i dont mind layering the wispys on after with another render pass.

many good examples are here in the volumetric methods in visual effects course from siggraph 2010, notice the almost displaced look of the density... http://magnuswrenninge.com/volumetricmethods

the most important thing i get from those courses is the 'modeling' of the volumes. they talk about using curves,patches, polygons, points/particles, level sets etc. we only have one way and that is with points/particles. of course they aren't using mental ray, most of them use a proprietary volume renderer.

as i said i have had some good luck using a dense mesh emitter that uses ICE and hand modeling for my overall cloud shape, then doing stuff like using weightmaps and turbulence to modify the positions/size of particles. feeding that per particle attributes into the fractal/cell shaders. but when it comes down to it that extra detail is always lost with the volume shader, it just always gets flattened or blurry. so i compensate by lowering the cell size and increasing memory but it gets way too slow without much return in detail.

i wonder what 3delight has for volumetrics in softimage now.

steven

Schoenberger

unread,
Feb 19, 2011, 8:34:41 AM2/19/11
to soft...@listproc.autodesk.com
More noise examples attached. It was actually a look dev for clouds.
It is a poly mesh with fractal and or cell shader.
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 


From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Vincent Fortin
Sent: Saturday, February 19, 2011 6:10 AM
Wolke_A01.1.jpg
Wolke_volB01.1.jpg
Wolke_volC01.1.jpg
Wolke_volD01.1.jpg
Wolke_volD01_soft.1.jpg
Wolke_volE01.1.jpg
Wolke_volF01.1.jpg
Wolke_volG01.1.jpg
Wolke_volG01_grundform.jpg

Schoenberger

unread,
Feb 19, 2011, 9:12:15 AM2/19/11
to soft...@listproc.autodesk.com
 
One issue if you use particles is that you can have a problem with overlapping of multiple particles.
If you mix two particles with a fine edge structure, then it gets blurry.
 
So you definitely need big particles (set the emission shape to sphere, make the alpha 0.85 for better viewport visualisation)
 
If your particles do not move, then to get rid of the "mixing particles with different noise":
I recommend that you add the noise AFTER the particle_density.
Get a multi math node.
Plug in the particle density as base, a cell as second (subtract) and then a fractal as third (subtract).
 
 
> the addition of a voxel framework in it would potentially blow the competition away
The volume shader is a kind of voxel shader.
 
> notice the almost displaced look of the density... http://magnuswrenninge.com/volumetricmethods
About the volumetrics siggrpah course 2010, most of the stuff is possible with the volume shader.
One thing missing is direct fluid conenction (That will perhaps come in a month).
And a better memory manegement for finer detail  (at the cost of render speed, will be implemented in v4.0 in half a year).
 
But for example the flow fields (Rhythm & Hues). You can already distort the volume in SI to get fluid-like edge noise.
Only issue is that they have an inhouse-tool to apply real fluid motion. But for non-fast-moving clouds you can apply a simple fractal:
This is for example a round displacement with a perlin noise:
But you can also apply other noise displacement for a more cloudy look.
 
 
The issue is not the shader, the issue is how to texture it. How to apply the right noise.
It is not easy/possible to create simple presets (there are two in the particle_shaper),
as the overall look changes with the particle sim (particles large<>small, many<>less, movement)
 
 
From the houdini video, I would say only the first one is not that easy.
Because of the large view and you have to somehow model the shape.
Same issue as my examples: To noisy edges to be real clouds. And the edge is not a round fluid style.
Main point: the clouds do not move. If they do not move, then your lighting can not flicker. Very Very big advantage.
 
I think I have to create some examples right now...
I will not model clouds with polygons (which would be way easier) as most of you do not have the BA_meshdistance shader.
I will try to model some clouds with particles.
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 

Sent: Saturday, February 19, 2011 6:42 AM

Schoenberger

unread,
Feb 19, 2011, 9:39:18 AM2/19/11
to soft...@listproc.autodesk.com
 
In the particle_density shader you have a bias for the falloff, increase it.
Set the fallof mode to linear.
If you really want that no density changes until half of the distance from the center, 
set the density shape intensity to 0.5 and set the density limit to 0.5.
Then it clamps the density.
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 

Sent: Saturday, February 19, 2011 1:55 AM

Dan Yargici

unread,
Feb 19, 2011, 9:37:16 AM2/19/11
to soft...@listproc.autodesk.com
> the addition of a voxel framework in it would potentially blow the competition away
The volume shader is a kind of voxel shader.

I'm guessing Steven is talking about fluid voxels type simulation framework?

I'm using Blender for a shot right now because the smoke stuff can do in minutes what would take days with em_Fluid and XSI's volume shaders (no offence to either yourself or Eric - I do still have constant use for your shaders and tools as well.... :)

I created the attached test in about 10 to 15 minutes and rendered 50 frames at half HD in 7 minutes on a crappy old workstation from 5 years ago.

It's worth crawling through the pool of broken glass that is Blender's interface and workflow to get the results.  Imagine if it was do-able in XSI.....


 
smoke_test.rar

Steven Caron

unread,
Feb 19, 2011, 2:17:02 PM2/19/11
to soft...@listproc.autodesk.com
actually that was vincent's comment. i am not talking about the simulation or motion, but rather rendering. also i am aware that the volume shader uses voxels (hence the cell parameters)

Steven Caron

unread,
Feb 19, 2011, 2:38:24 PM2/19/11
to soft...@listproc.autodesk.com
hey holger,

the sections in the document about fluids is not what i was trying to share, but rather just the parts about clouds and the many options to model the density for the renderer. of course sending emfluid or slipstream data straight into the shader for direct volume render is desirable!

also i understand they are all using proprietary renderers and simulators, i was trying to point out what i am trying to achieve.


The issue is not the shader, the issue is how to texture it. How to apply the right noise.
It is not easy/possible to create simple presets (there are two in the particle_shaper),
as the overall look changes with the particle sim (particles large<>small, many<>less, movement)

i dont need a preset, nor am i requesting one. if anything more documentation so i can understand all the ways i can apply/control the noise. i also agree the look changes with the particle sim, and i have been using the size attribute in the render tree and have been getting better results.


And a better memory manegement for finer detail  (at the cost of render speed, will be implemented in v4.0 in half a year).

this is good news! i want to make the cell sizes really small to get that detail and i hit a render time wall.


I think I have to create some examples right now...
I will not model clouds with polygons (which would be way easier) as most of you do not have the BA_meshdistance shader.
I will try to model some clouds with particles.

i will be happy to join you in doing so, just as long as we talk about how it was done. ie. documentation

steven

Schoenberger

unread,
Feb 19, 2011, 3:16:51 PM2/19/11
to soft...@listproc.autodesk.com
Hmm, seems that my last mail was not transmitted. Trying without attachment...
 
This is a simple scene with particles,
the fractal applied after the particles
and increased contrast.
The render has a comped depth fog (depth rendered into framebuffer)
 
 
@Steven:
Try to use the contrast slider in the volume shader, I think that will help you more in your case than changing the particle falloff.
 
And I think I will find some other way to create what you described. 
There has to be a way how to texture the clouds differently. 
Would probably be slower in rendering again.
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 

Sent: Saturday, February 19, 2011 8:38 PM

Steven Caron

unread,
Feb 19, 2011, 3:50:20 PM2/19/11
to soft...@listproc.autodesk.com
that is nice! the larger scale is working better here. but i think the detail could be a better, i hope its something the future memory management changes will improve?

i attached a wip, there are aspects i like about it and others i dont. i have yet to implement your suggestions yet. which if i do i think i will start from scratch so i can try and understand it better.

s
backlit_cloud_2.jpg

Steven Caron

unread,
Feb 19, 2011, 5:15:29 PM2/19/11
to soft...@listproc.autodesk.com
hey holger,

it would be nice to control that 'falloff bias' per particle, any reason this wasn't exposed?

s

Schoenberger

unread,
Feb 19, 2011, 5:33:06 PM2/19/11
to soft...@listproc.autodesk.com
 
>it would be nice to control that 'falloff bias' per particle, any reason this wasn't exposed?
Speed. But as you have seen in my other comments about a new version, the next version will be slower.
So perhaps I can also add that. (also, I can perhaps re-add the falloff-start, which is only available if you enable bubble density, it was removed as to many parameter confuse artists to much)
Also, are your particles very small? If you load my scene, I am using bigger ones that the structure/texture does not mix/average/overlap.
 
>that is nice! the larger scale is working better here. but i think the detail could be a better, i hope its something the future memory management changes will improve?
Hmm, your render does not seem to take a lot of memory, even with small cells. But for that I would need to see a render with cell interpoaltion "none".
Clouds have a lot or scattering, so usally you can only see detail on the edge. Which is not influenced by the cell size (lighting/shading).
If it would not be water clouds, more pyroclastic dust, you could enable the shading with normal enabled. That gives you more detail.
 
 
If I take a look at your egde of the cloud, it looks natural. The inner parts are more spotty.
 
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 

Sent: Saturday, February 19, 2011 11:15 PM

To: soft...@listproc.autodesk.com
Subject: Re: Clouds - Simul software technology

Schoenberger

unread,
Feb 19, 2011, 6:59:53 PM2/19/11
to soft...@listproc.autodesk.com
About the displacing and modelling.
With the BA volume addon you get the mesh distance to model with polygons.
 
But I forgot to mention one thing:
If you take a look at the paper, what they describe as "displacement" (not the vector displacement, that it something similar to what you have now).
The main difference are the texture coordinates. (Also something of the next version)
Spherical texture coordinates, see attached image (turbulence/vein as noise).
 
I have tried to replicate it with the attached compound.
You have to disable "unique texture coordinates" in the particle shader.
Then use the spherical texture coordinate for the fractal or cell that you use for the density shape.
Way esier to get a shape with that.
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 

Sent: Saturday, February 19, 2011 6:42 AM
BA_PT_Texcoord_spherical.xsishaders
BA_Textcoord_spherical.jpg

Steven Caron

unread,
Feb 19, 2011, 7:11:50 PM2/19/11
to soft...@listproc.autodesk.com
thanks holger

the dialogue we are having is giving me more confidence in the volume shader's flexibility.

you mention the BA Volume addon comes with mesh distance? is this currently available replacement/upgrade for the softimage volume shader?

thanks for the example shader, i will investigate with it. currently i am getting acceptable results for the work i am doing but would like to push it further in the future.

s

Schoenberger

unread,
Feb 19, 2011, 7:26:58 PM2/19/11
to soft...@listproc.autodesk.com
 
 >BA Volume addon comes with mesh distance? is this currently available replacement/upgrade for the softimage volume shader?g
It is an addon with more shaders. The volume shader itself is the same version (so not included in the XSI package).
 
And another example file:
Turbulence with sperical coord:
Added a second turbulence with default 3D coord:
Scene:
 
 
Ok, enough samples, should do less samples with the old version and better work on the next shader (fluids). :-)
 
And I think I will do more basic sample scene for the next version 4.0 of the volume shader .
But that will take a while...(half a year probably)
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 

Sent: Sunday, February 20, 2011 1:12 AM

Steven Caron

unread,
Feb 19, 2011, 7:38:14 PM2/19/11
to soft...@listproc.autodesk.com
thanks for the link, the mesh distance shader has been around a while i see. i need to give that a go.

again, this dialogue is great! and i hope can help improve the version 4 of the volume shader. with that said i think many many example scenes would be best. from simple isolated examples to production level. if you need beta testers to help generate these samples let me know ;)

s

Vincent Fortin

unread,
Feb 19, 2011, 10:17:44 PM2/19/11
to soft...@listproc.autodesk.com
Actually I'm taking about voxels as a whole. It is possible to operate on a 3d grid like you do with points in ICE. Gradients, fractals, math operations, procedural displacement of density, optimization, caching, all of this can happen before rendering, with visual feedback. So it's definitely not just for sims. Of course things like fluid dynamics, collisions, curl noise can relatively easily be implemented by third parties once the level set framework is in place. It's an industry standard and ICE is screaming for it. I'm just not sure if this it's on Softimage's todo list right now. Eric Mootz has implemented his own structure for the needs of his tools and I'm positive that somewhere lies a good opportunity for him to open up his code and expose a few basic nodes for ICE. If that was even possible we would see Softimage raise to a higher level... especially that we now know from official source that Holger and Mootz are working together to get that data rendered out smoothly :-) Thanks Holger for the nice examples in this thread, will try those ASAP!
Vincent
Some food for thought -last Houdini links I promise-:

Schoenberger

unread,
Feb 20, 2011, 7:12:01 AM2/20/11
to soft...@listproc.autodesk.com
 
Why operate on a 3D grid when you can do the same math operations without a grid?
If you have a shader that gets the density to an object, it does not matter if you put it into a grid or directly render it.
Ok, there are some things you can not do without a grid. For example a blur filter.
 
But most of these voxel-shaders have an issue with the detail.
Either you create a voxel per pixel or your voxel computations are blurred.
And just try to set the cell size of the current volume shader to the size of a pixel.
Frustrum shaped buffers can help, afaik Mantra can use them at render time, but you get issues if you have reflections in your scene.
 
 
With the current implementation:
You can convert an object to a volume with the meshdistance shader.
Math operations.. see my examples I send. I am using the multi math node.
(You already had the possibility to use math operations,
but I do not know how many have acutally used it, although it is already used in the default particle shaper compound)
Displacement: See my last example, the texture coordinate was the main issue.
 
 
Visual feedback:
That is not that easy.
Cause on one side you have the mental ray render tree, on the other side you have a viewport.
You would need to convert all shaders in the render tree. You cannot use 3rd party texture shaders.
In the end: Nothing I can do about if the user wants to have to freedom to use all MRay shaders and not a closed volume plugin.
It would perhaps be possible to render a frame and then keep the voxels of that frame for the viewport.
It would give you the same look as if you enable "render lookup table only" in the preview mode.
 
For fast feedback: You have the preview tab in the volume shader.
Fast feedback for viewport, final quality for batch rendering.
 
 
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 


From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Vincent Fortin
Sent: Sunday, February 20, 2011 4:18 AM

To: soft...@listproc.autodesk.com
Subject: Re: Clouds - Simul software technology

Ben Beckett

unread,
Feb 21, 2011, 9:18:43 AM2/21/11
to soft...@listproc.autodesk.com
I wish this tread happend 6 month ago. Brill

Steven Caron

unread,
Feb 22, 2011, 7:33:10 PM2/22/11
to soft...@listproc.autodesk.com
hey holger,

so votch and i have been pushing our clouds, separately but both effectively. today he shared with me his tests from the weekend which looked very nice! but we both had a question which i would hope you could clear up.

what is the difference between using the scalar multi math and fractal after the particle density, vs just putting the various fractals into the density port of the particle density shader? this seems to get the result we want more directly and easier. is the texture coordinates or scale of the fractal some how different when used this way?

steven

peter boeykens

unread,
Feb 23, 2011, 3:14:50 AM2/23/11
to soft...@listproc.autodesk.com
afaik (but Holger is the man of course and he will probably correct me)...
 
if you put the fractal into the density inputs, you are getting per-particle texture coordinates,
so that each particle can have its own texture coordinates that stick with it (each individual particle looks "unique")
 
if you put the fractal after the particle density shader, you are getting a texture in world space that the particles swim through.
This is useful when you are not using particles, but rather a solid volume to do clouds.
eg. a huge cube is rendered as a volume from which you substract a few fractals, for instance first a large low detail one to get the overal form of the clouds, and then a smaller high detail one to get the smaller structures.

Schoenberger

unread,
Feb 23, 2011, 2:56:36 PM2/23/11
to soft...@listproc.autodesk.com
Right.
 
If you plug it into the particles, the texture moves/rotates/scales with the particle.
But it has to be evaluated for each particle, so slower if you have many overlapping particles.
And the texture averages if particles overlap.
 
The other one can be compared to a laser (particle) in a club that moves through the fog (fractal).
There is no mixing/averaging of the fractals if you have multiple particles overlapping.
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 
 


From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of peter boeykens
Sent: Wednesday, February 23, 2011 9:15 AM

Steven Caron

unread,
Feb 23, 2011, 2:58:27 PM2/23/11
to soft...@listproc.autodesk.com
ok, it looked as peter suggested, just wanted to be sure.

s

Len Krenzler

unread,
Feb 28, 2011, 12:46:37 PM2/28/11
to soft...@listproc.autodesk.com
I've been trying to figure out how to do this (put the fractal after the density shader).  Does anyone have a screenshot perhaps?
-- 
_________________________________________________

Len Krenzler - Creative Control Media Productions

Phone: 780.463.3126

www.creativecontrol.ca - l...@creativecontrol.ca

Schoenberger

unread,
Feb 28, 2011, 3:02:44 PM2/28/11
to soft...@listproc.autodesk.com
I have send some examples in this thread.
Some as attachment, some as link.
I think there was one with that setup.
 
Otherwise, it is very simple.
Instead of pluggin the particle_density direclty into the volume, plug in in a multi math first.
 
 
Holger Schönberger
technical director
The day has 24 hours, if that does not suffice, I will take the night
 

 
 


From: softimag...@listproc.autodesk.com [mailto:softimag...@listproc.autodesk.com] On Behalf Of Len Krenzler
Sent: Monday, February 28, 2011 6:47 PM

peter boeykens

unread,
Feb 28, 2011, 3:54:53 PM2/28/11
to soft...@listproc.autodesk.com
A
> C - D
B

four nodes:
particle_density (A), fractal_scalar(B), scalar_basic ( C) , particle_volume
(D).

particle_density's density output goes to the first input on the
scalar_basic
fractal_scalar's output goes to the second input on the scalar_basic.
scalar_basic goes to the (density) input on the volume_shader.
set the basic to subtract.

like that, the density of the particles gets "cut out" with the fractal
before it goes into the volume.

plenty of variations possible.
You can add, subtract, mix, you can daisy chain or blend several inputs,
you can put fractals into fractals into fractals before using them,...
you can do a separate tree for the color and for the density and for the
ambient of the volume -
the possibilities are endless.

From: Len Krenzler
Sent: Monday, February 28, 2011 6:46 PM
To: soft...@listproc.autodesk.com
Subject: Re: Clouds - Simul software technology