multi-displacements bug

   8150   12   0
User Avatar
Member
243 posts
Joined: Oct. 2007
Offline
Hi!
I'm experimenting a multi-displacement shader and have little problems.
There are some artifacts in my renders and I have no clue what this might be. Can somebody have a look to help me please? I'm not an expert at all in shaders so don't be shy to explain well in details.

thanks very much!

Attachments:
displacement_bug.jpg (87.5 KB)
multidisplace_test_001.rar (30.6 KB)

JR Gauthier
Character Animation & Design
www.turboatomic.com
http://www.vimeo.com/user2847970 [vimeo.com]
User Avatar
Member
349 posts
Joined: Aug. 2006
Offline
Go to the Object level of the geometry that is displaced, and hit the gear icon to edit the parameter interface (edit rendering parameters). Then go to…

Mantra 10—>Shading—>and select Displacement bound (vm displacebound) and Displacement space (vm displacespace) and move them across to the existing parameters dialog, and apply and accept.

Go to the rendering folder on the goemetry container, and at the bottom change the displacement bound to somthing small like 0.1. Have a fiddle with this setting to get things less artifact-y.
User Avatar
Member
349 posts
Joined: Aug. 2006
Offline
Or add a property node inside shops, and then do the edit parameter interface technique above. but in the property node, and add it to the output shader.
User Avatar
Member
243 posts
Joined: Oct. 2007
Offline
Hi Ragupasta,
thanks for the quick answer. Have you actually tryed it in my scene(could you post it please?)
I already have displacement bound properties in my materiel properties. And I played with it's value and doesn't make a difference in any cases. Maybe there's is something else I'm doing wrong..
But I have a feeling it's some texture value going out of control.
Even when I check bump only in my displacements node, I still have some weird artifacts at render(I'm under the impression that displacement bound is only valid when there are displacements involved, maybe I'm wrong)…I included another image…

Hope you know what it is…

Thanks alot!
Edited by - Nov. 23, 2009 07:22:10

Attachments:
bump_only.jpg (119.1 KB)

JR Gauthier
Character Animation & Design
www.turboatomic.com
http://www.vimeo.com/user2847970 [vimeo.com]
User Avatar
Member
349 posts
Joined: Aug. 2006
Offline
Has a quick look over your file, and it renders ok for me. Try deleting the shadow map .rat file and make a new one.

Attachments:
disp.jpg (40.5 KB)

User Avatar
Member
243 posts
Joined: Oct. 2007
Offline
Doesn't work for me..
However, less artifact appear if the dicing of the grid is low(setted to 1 for exemple). Did you try a dicing of 3 for any chances, if it still works on your side I will be totally confused.

Thanks man! I really appreciate your help!
JR Gauthier
Character Animation & Design
www.turboatomic.com
http://www.vimeo.com/user2847970 [vimeo.com]
User Avatar
Member
349 posts
Joined: Aug. 2006
Offline
Dicing on the grid object is:

Shading quality: 3
Ray Shading quality: 3

Renders fine for me.
User Avatar
Member
243 posts
Joined: Oct. 2007
Offline
Hmmm what version of Houdini are you using?
I notice a difference between our 2 renders thought, yours seems to have much less fine details in the displacement.

I've done a test using the render view and even thought the antialiasing, is not perfect, I have much less artifacts…I don't understand what's going on…
:shock:

Attachments:
ipr.jpg (95.9 KB)

JR Gauthier
Character Animation & Design
www.turboatomic.com
http://www.vimeo.com/user2847970 [vimeo.com]
User Avatar
Member
349 posts
Joined: Aug. 2006
Offline
Im using 10.0.430
User Avatar
Member
89 posts
Joined: April 2008
Offline
I think the problem you're having comes from chaining together the normals from the “displace along normal” nodes. Even if the problem is not as bad in raytrace (ipr), you're still going to get artifacts from distorted normals. Just use the normal output from the last node in the chain, which internally calls computenormal(), and all should be fine.

Another thing is that while there's nothing wrong with using multiple “displace along normal” nodes, you can get the same results more efficiently by adding together the nodes you're using as the “amount” inputs after they're multiplied by a scale factor (your “scale” values in the displace nodes) and feeding that into a single displace along normal node.

hth
User Avatar
Member
89 posts
Joined: April 2008
Offline
Oh, I just realized that you're chaining together the normals to get non-linear displacements?
If that's the case, then still, I dont think computenormal() is going to give you what you want.
Maybe this would help?

http://www.renderman.org/RMR/Shaders/IDShaders/IDgloop.sl [renderman.org]
User Avatar
Member
12508 posts
Joined: July 2005
Online
Have you added and engaged “Re-dice Displacements”? For MP renders, this might help you a bit.
Jason Iversen, Technology Supervisor & FX Pipeline/R+D Lead @ Weta FX
also, http://www.odforce.net [www.odforce.net]
User Avatar
Member
243 posts
Joined: Oct. 2007
Offline
Thanks for the replies guys!

mrice: yeah, preferably I'd like the displacements to follow normals of the previous displacement. I will still give a try without branching 3 displace along normal together to see the difference. And I'll have a look at the link you sent me…I hope I will understand a word or 2…Just read it, and as I suspected, I don't understand a thing..can someone translate it in english please?

Here's the network + my displacement properties, so people can see it right away. I already have re-dice displacements activated.

Thanks again, I really hope i'll find a solution to this. I'm addicted to displacement map now, thanks to Houdini.

Attachments:
network.jpg (122.7 KB)

JR Gauthier
Character Animation & Design
www.turboatomic.com
http://www.vimeo.com/user2847970 [vimeo.com]
  • Quick Links