Page 46 of 92
Re: BlendLuxCore Development
Posted: Mon Mar 11, 2019 2:51 pm
by B.Y.O.B.
Dade wrote: ↑Mon Mar 11, 2019 2:38 pm
Well, do the Dot() inside the texture if you have but don't add a field to HitPoint object (i.e. GPU memory usage).
I don't need the extra field on the GPU because Bidir doesn't work on the GPU (I can just use fixedDir there because it's always the camera vector).
Dade wrote: ↑Mon Mar 11, 2019 2:38 pm
how you are supposed to know the camera dir while tracing paths from the light sources (i.e. BiDir, PhotonGI, etc.) ?
I thought I could maybe calculate
cameraDir = camera.position - hitPoint.p;
Can I check the camera position when initializing hitPoint?
Re: BlendLuxCore Development
Posted: Mon Mar 11, 2019 3:02 pm
by Dade
B.Y.O.B. wrote: ↑Mon Mar 11, 2019 2:51 pm
I thought I could maybe calculate
cameraDir = camera.position - hitPoint.p;
Can I check the camera position when initializing hitPoint?
Uh ? Can you recap of what we are talking about ? I have lost you.
An HitPoint is the result of a ray hitting a surface or of a volume scattering event. Most of the times a ray doesn't come from a camera (i.e it comes form a light source, another hit point, another volume scattering event, etc.).
Re: BlendLuxCore Development
Posted: Mon Mar 11, 2019 4:29 pm
by B.Y.O.B.
Yes, I was wrong, taking the camera dir is bullshit.
I think this kind of texture can only work in a unidirectional pathtracer, as you said it makes no sense for a lighttracer.
In my opinion this means it shouldn't be added at all, inconsistency between engines is bad.
@Sharlybg the dust effect you mentioned is possible without this by the way, for that you only need Dot(hitPoint.shadeN, Vector(0, 0, 1)).
Here is the result of
Code: Select all
Dot(hitPoint.shadeN, hitPoint.fixedDir)
for pathcpu, lightcpu, bidircpu and cycles:
Re: BlendLuxCore Development
Posted: Tue Mar 12, 2019 10:38 am
by Sharlybg
It doesn't work on light cpu. so we are not going to be able to tweak this viewport render ? how reproduce it in blendluxcore node system
Re: BlendLuxCore Development
Posted: Sun Mar 24, 2019 10:05 pm
by B.Y.O.B.
Sharlybg wrote: ↑Mon Mar 04, 2019 10:15 am
I guess for dust you would need the world Z coordinate as input instead of the camera direction? So an arbitrary normal input like in Cycles' fresnel node would be good.
I'm not a great expert about the tech behind such kind of things. I mostly don't like too complexe shader node to achieve basic natural effect.One of the reason i don't like cycles node system and node tricks.I think tools should allow artist to concentrate on art . So my point is that if you can yourself produce this effect (the principle) with luxcore node system and provide us the how to do i'm ok
The dust effect you wanted is now possible in v2.2alpha1.
You will need two of the new nodes from the "Utils" category: Hitpoint Info and Dot Product.
Then you can create a setup like in the attached images.
(Rhino model from
http://threedscans.com/depot-des-sculpt ... ris/rhino/)
Re: BlendLuxCore Development
Posted: Sun Mar 24, 2019 10:32 pm
by B.Y.O.B.
Played around a bit more. You can control the falloff of the dot product with the new "power" option in the math node.
Re: BlendLuxCore Development
Posted: Mon Mar 25, 2019 6:23 am
by provisory
Re: BlendLuxCore Development
Posted: Mon Mar 25, 2019 7:05 am
by lacilaci
Dust looks nice but I think a proper AO/dirt map with controls over distance and mapping(for variation) is needed for better dirt distribution (also with possibility to select self contact only or the whole scene)
Re: BlendLuxCore Development
Posted: Mon Mar 25, 2019 8:42 am
by epilectrolytics
Thanks for the new nodes/functions, much appreciated!
Re: BlendLuxCore Development
Posted: Mon Mar 25, 2019 10:07 am
by Sharlybg
Like you , you know ?
thanks you very much !