OctaneRender™ 2020.2 RC7 [obsolete]

[Kalua]

When scattering an object, is there a way to avoid objects being scattered in the same position?
Here are about 4 objects (ferns) scattered in the same place.
-

scattering.JPG

Which distribution model did you use? Surface area and Relative density should produce better scattered objects. If you use one of these and you still have 4 on the same place, maybe there is a bug somewhere, in which case it would be good to have the scene so we can investigate.

Does topology quality influence scatter distribution/concentration?

[jobigoud]

Does topology quality influence scatter distribution/concentration?

It depends on the distribution method.
- If you choose "Surface area" it will take into account the relative area of the polygons so it should generate instances uniformly over the surface independently of the topology, never multiple instances in the same place with empty space around like that.
- If you choose "Relative density" it will draw random locations in UV space, so maybe if some polygons are stretched/squeezed in UV space compared to world space it might result in unexpected distribution.
- If you choose "Vertex" and your polygons vertices are not welded together, it won't weld them, so you might get this kind of result if there are multiple vertices in the same location.
- If you choose "Polygon center" it will also of course be impacted by the topology.

There is no push-apart or collision radius to isolate instances from each other at the moment, so you will get intersecting instances if the density is high enough. It shouldn't be as extreme as in the screenshot though, especially when there is empty space around.

edit to add: for "Relative density", it also doesn't work well when there are overlapping triangles in UV space.

[blockmind]

Hello, when will the stable version of 2020.2 be released? Still we got no release to run with Nvidia 30xx series cards to use denoising effectively. Old versions doesn't support denoising. This one has random timing on denoising to complete. These cards are in the market for months. Is there any ETA please?

[Zay]

Which distribution model did you use? Surface area and Relative density should produce better scattered objects. If you use one of these and you still have 4 on the same place, maybe there is a bug somewhere, in which case it would be good to have the scene so we can investigate.

I use a surface (plane) with a relative density map. But I think the problem here is, if the painted map area is too narrow, you have to crank the density way up to cover the area and then it starts to populate in the same place or very close to it.

I think adding a slider for "even space" will fix it.

Here is the bottom of a plane with trees scattered. Some are way to close or even intersecting. I will send a PM with the test file.
-

[abstrax]

Hello, when will the stable version of 2020.2 be released? Still we got no release to run with Nvidia 30xx series cards to use denoising effectively. Old versions doesn't support denoising. This one has random timing on denoising to complete. These cards are in the market for months. Is there any ETA please?

Good question. We are getting very close. We made a 2020.2 release build this weekend, but will have to do another one as various issues have come up since then. And we are also waiting for some plugins to get ready. So the current plan is end of this week, but that is subject to change.

[blockmind]

Hello, when will the stable version of 2020.2 be released? Still we got no release to run with Nvidia 30xx series cards to use denoising effectively. Old versions doesn't support denoising. This one has random timing on denoising to complete. These cards are in the market for months. Is there any ETA please?

Good question. We are getting very close. We made a 2020.2 release build this weekend, but will have to do another one as various issues have come up since then. And we are also waiting for some plugins to get ready. So the current plan is end of this week, but that is subject to change.

Thanks for your reply. I'm sure you're struggling with so much issues. Good luck on them.

"I can't use Octane for months". I'm sure this just sums up my pain

[funk]

BUG: Some compositing "Info passes normalized" don't work

It looks like a few of the "info passes normalized" aren't working as expected.

1. Load orbx and select render target
2. Select the UV pass to see what it looks like, then select the Out1 pass

RESULT: The out1 pass is set to UV Coordinates normalized, yet looks nothing like the UV pass

Further steps:

3. Change the "Layer 1" input passes to check the rest of the "info passes normalized" passes.

RESULT: The following look incorrect - Uv coord normalized, motion vector normalized, texture tangent normalized
I think everything after the z-depth pass shows the same output.

[funk]

BUG: Render AOV output > Info passes > Trace Time doesnt exist

There is a pass listed as Render AOV output > Info passes > Trace Time. I can't see any place to enable this pass though.

Code: Select all

Started logging on 04.02.21 12:41:40

OctaneRender Enterprise 2020.2 RC7 (10021100)

Trace time render pass has not been enabled, the compositing of a composite AOV could yeild an unexpected result

[Daniel_Ward]

Hi Jobigoud,

I could probably work with culling and source selection as a limitation for surface scattering on meshes.
How far away do you think you are for a fix on the UV's outside the unit square issue?

Cheers, Dan.

Hi,
Because these tools work on the CPU they don't have access to the texturing pipeline of Octane.
We will have a fix for a subset of cases: when the mesh has UVs outside the unit square and the texture is mapped with a transform. This is only for the culling map, the source object selection map and the transforms maps and only when the input is an image texture. It will not be supported for the relative density map due to its different approach.

I'll try to document the limitations of the tool here:

Geometry
- Texture and Vertex displacement are ignored when computing the instance positions.
- Normal maps are ignored when computing the instance normal (only geometric and smooth normals are supported).
- If the input surface has a non-uniform scaling factor the distribution by area will have a bias.
- If the input surface has a skewing transform the instances placed on it will also be skewed and their normals will be wrong.

UV mapping
- UV mapping for culling and transform textures: the projection is fixed to Mesh UV, using the first UV set and the border mode is fixed to wrap around.
- UV transforms won't work with procedural textures as input.
- UDIM/Image tiles won't work.
- UV mapping for the relative density texture: the UV transform is not supported.

[jobigoud]

How far away do you think you are for a fix on the UV's outside the unit square issue?

There is a change in to support it for the culling and transform textures but as it was made very close to 2020.2 final release without going through the beta or RC cycle we preferred to delay its inclusion until the first 2021 beta version.