PDA

View Full Version : Show your support for Subpixel displacement



hrgiger
09-27-2005, 11:09 AM
Now that 8.5 and 9 are in their final stages (hopefully), I'd really like to get the community rallying behind a push for sub-pixel displacement. I've thought about getting zbrush but I've heard that the results once a model is painted with displacment and brought back into Lightwave, the results aren't the greatest.
Sub-pixel displacement pretty please. With sugar on top.

Captain Obvious
09-27-2005, 11:15 AM
Lightwave 9 is supposed to have adaptive subdivision. This is pretty much the same thing as micropoly or subpixel displacement.

http://www.newtek.com/lightwave/lw9featureprelim.php
"Adaptive subdivision of a mesh based on distance from camera and visibility"

hrgiger
09-27-2005, 11:21 AM
Not the same thing at all. There are also no tools for painting displacements directly on the model. I'm talking about using normal maps to generate high resolution detail from lower resolution models. Like in Zbrush or now modo.

BazC
09-28-2005, 04:26 AM
There seems to be some confusion here. Normal maps have nothing to do with displacement, they are more or less sophisticated bump maps. I believe LW9 will support them

Lightwave's adaptive subdivision is also nothing like sub-pixel displacement as I understand it. Lightwave will subdivide geometry intelligently at render time. It's an improvement over the current system but still has the potential to add significantly to render times.

True sub-pixel displacement is a rendering effect, it looks exactly the same but doesn't involve real subdivision of geometry and it adds very little to render times no matter how fine the detail. It's used in PRman and other Renderman compliant renderers, also Poser 5 and 6 (!)

I think (but I'm not sure) that Daz Studio (using 3Delight as it's renderer) also uses this technique and it's free if you want to play with true sub-pixel displacement. www.daz3d.com - Baz

hrgiger
09-28-2005, 05:07 AM
My mistake. I guess I thought that normal maps were involved in the displacement process. I did some research and found a good webpage which explains what normal maps are and how to use them here: http://www.monitorstudios.com/bcloward/tutorials_normal_maps1.html

I knew however that adaptive subdivsion had nothing to do with sub-pixel displacement and had feared that it would add to rendering times.

I really just want to have the ability to use displacement maps like in zbrush or at least have lightwave support sub-pixel displacements so that models made in zbrush could be used in Lightwave without the loss of detail.

BazC
09-28-2005, 05:13 AM
I'm guessing you will though it will involve a little more work. You should be able to add medium res detail using LWs new adaptive subdivision with minimal increase in render times This will take care of visible changes to the sillouette. Super fine detail can then be added with a normal map, again with minimal effect on render times. The result should look pretty much like subpixel displacement!

Of course we won't know for sure till we get to play with LW9!

Captain Obvious
09-28-2005, 05:42 AM
Lightwave's adaptive subdivision is also nothing like sub-pixel displacement as I understand it. Lightwave will subdivide geometry intelligently at render time. It's an improvement over the current system but still has the potential to add significantly to render times.
I know they're not technically the same thing, but the results are fairly similar, and I think that's what matters most.

pixelinfected
09-28-2005, 06:34 AM
normal map are something like superbump, and are avaible like free plugin like tb_normal.p

from zbrush you can out dispacement map to deform geometry, and normal map for fine dectail. try to render high dectail with true geometry it's very cpu consuming, and not every times is good thing,

newtek integrate adaptive subpatch displacement like they show, that are a good thing, be cause integrate quality and fast render at same times.

BazC
09-28-2005, 06:41 AM
I know they're not technically the same thing, but the results are fairly similar, and I think that's what matters most.


True to a point! It really depends how efficient Newtek's implementation is, true Subpixel displacement is reputedly MUCH faster. As a matter of interest, Maxon added Subpolygon displacement to Cinema R9, it's faster than the older displacement system but it can still ad a lot of time to renders. I'd think carefully before using it but then I have a puny 1.25GHz G4 lol!

BazC
09-28-2005, 06:44 AM
As a sidenote, I'd love to see subpixel displacement in Lightwave but from what I've read true subpixel displacement doesn't play nice with raytracers. They've managed to add raytracing to PRman so presumably Pixar have cracked the problem but I'm not sure if anyone else has?

pixelinfected
09-28-2005, 06:52 AM
i use renderman for some years, but renderman is mainly a no raytracer render, at pixar add raytracing function, but usually they use many tricks to have same result or raytracing with very low rendertime, to optimize high res rendering.
Subpixeldisplacement have many good things, but have a bad thing, until you haven't render it, you cannot see the displacement, that mean some thing can compenetrate other obj. if you use adaptive subpatch displacement you can see in teoric real time the displacement, than help you to check collision and many other things that you need to know during complex animation. good thing of subpixeldisplacement is show in very fine dectails, but actually normal map can work fine, and are less cpu consuming.

BazC
09-28-2005, 07:02 AM
Subpixeldisplacement have many good things, but have a bad thing, until you haven't render it, you cannot see the displacement, that mean some thing can compenetrate other obj. if you use adaptive subpatch displacement you can see in teoric real time the displacement, than help you to check collision and many other things that you need to know during complex animation.


Good point! I hadn't considered that aspect :)

pixelinfected
09-28-2005, 07:07 AM
when i used renderman subpixeldisplacement seems me a miracle, lw was at version 3.5, which have displacement, but under amiga4000, and with 16 mega of ram, you cannot build a so complex mesh to reproduce subpixeldisplacement....

old times, renderman on a 100 mhz pentium, with 32 mb of ram, and it was a big number at that times, i use it with nextstep os, and old but good unix os, which the old grand father of actual mac os X.

at that times, the problem is small, actually could be more important.
have a niceday.

Karmacop
09-28-2005, 09:11 AM
As a sidenote, I'd love to see subpixel displacement in Lightwave but from what I've read true subpixel displacement doesn't play nice with raytracers. They've managed to add raytracing to PRman so presumably Pixar have cracked the problem but I'm not sure if anyone else has?

That's what I've heard too. I did some searching the other night though and I couldn't get a certain answer, but I think you can do it. Anyone with mental ray could try it out for us, because it does raytracing and sub pixel displacement.

JMarc
09-28-2005, 09:55 AM
The adaptive subdivision demos I saw at Siggraph were wicked fast renders with over a million polys. I was quite impressed with render times. Unless things have gotten worse since then, don't worry about how it affects render times. The new renderer seems to handle it all quite well.

pixelinfected
09-28-2005, 10:23 AM
i not express idea in correct way, is not a problem to build a render with raytracing and subpixel displacement, but the problem is to optimize render time.
pixar is famous for their famous fake
- reflection (reflection map builded from shiny obj)
- refraction (refraction shader to simulate it)
- shadow (deep shadow which see fine dectail and are faster to render)
- antialias (many different techniques to antialias section, from texture antialias to antialias pass on edge obj, and many postpro antialias tech)
- motion blur (different tech to rebuild a very good mb, without infinity sampling)
- nurbs and subpatch refining at pixel (in Renderman nurbs and subdivision surface are subdivided and refined at pixel, engine decide between size, distance and size on screen how many is the subdivision level must be applied at mesh in that part)

this last features seems like that newtek tell about lw9 features, i read in newtek site and in spinquad thread from ones of developer, that you can setup sampling for pixel for subdivision setup (if i understand correctly), but difference is that renderman setup that for render, lw is also in real displacement like you can see in thread, and that is better to work with mesh.

http://www.spinquad.com/forums/showthread.php?t=7607

Karmacop
09-28-2005, 11:37 AM
I've heard that Pixar didn't use any antialiasing or motion blur in the way that we know it. Each pixel is randomly giving a slice in time to render, which would make a render very pixelated as far as motionblur goes, but since it's film res they just blur the image a bit and it looks fine .. maybe I'm wrong though.

Yes that last feature does sound similar to LW9's but they are very different. LW9 subdivides the mesh based on the distance to camera etc, while renderman (using the REYES system) actually creates 4 polygons per pixel based on the scene. It's hard to explain in words, but trust me, they're different :)

pixelinfected
09-28-2005, 12:16 PM
HI Karmacop,
from what i read on new features of lw9, there is a features about how many subsample to do for every pixels which is similar to renderman features, but i know the tech and idea behind are deep different, but i'm interest to the result.

"This approach has certain other benefits, such as significantly improved displacement performance, as the mesh is highly optimized. As an example, if you are rendering a 640 by 480 image and you sent a sub-pixel value of 4 (aka 16 subs per pixel), then you simply get 16 polygons at max per pixel in the scene. This results in the ability to displace and deform with higher accuracy. While not true micro-poly displacement, this method will yield similar results in many cases. "

from newtek web part of lw9
i think is enought for a general render dectail, also be cause 16 sample for a single pixel is enought to have a very good quality on render.

i check on renderman companion and renderman for movie books, and i understand that you told.

GregMalick
09-28-2005, 01:21 PM
:agree:

This is an important feature.
NT Developers, please make sure to include it in LW9.

Nemoid
09-29-2005, 06:18 AM
I think that raytracing and subpixel displacement will be handled very well by apps like -cough cough- Modo i've seen the recent rendering made by Allen hastings in lux forums and i have to say that it rocks and render times can be optimized greatly through numnber of rays settings, bounces and more.

this being said, the newtek feature is a huge step in the right direction good for production, and i think real subpixel displacement will come for Lw as well ASAP :)