OctaneRender™ Standalone 3.07 TEST 6

[Daniel79]

so i have to invert the normal of the plane?
Thanks a lot

[Daniel79]

but there is another thing that is not ok but maybe is normal....whit a texture enviroment the plan whit shadow catcher enable is not trasparet...


I solved, with the power at 0.1 of enviroment the plane are visible whit power 1 not.

[abstrax]

but there is another thing that is not ok but maybe is normal....whit a texture enviroment the plan whit shadow catcher enable is not trasparet...


I solved, with the power at 0.1 of enviroment the plane are visible whit power 1 not.

You probably have an emitter in the building, which is also casting a shadow. Either disable its emitter or turn off "cast shadows" in the emitter.

[abstrax]

Thanks you all for your feedback regarding the noise in participating media. While there is no easy fix for it we will be looking into it. There are a few things you can do to at least improve the situation a little bit:

As a first step, so get a good view of the problem, I set the camera response to linear (most response curves increase the contrast and thus make the situation worse), the gamma to 2.2, vignette to 0, max samples to 20000, removed the cube transformation, rotated the distribution texture (so I can see it in the preview) and set the camera to look from the left:

This is the IES distribution texture you are using:

ies_distribution.png (38.48 KiB) Viewed 4295 times

Obviously, a cube is not a good shape for a directed emitter like this. A better choice would be a hemisphere:

or even better a simple quad:

I had to modify the emitter power for the various shapes to keep the overall emitter power in the same ballpark. Since the quad is best I'm using the quad emitter below.

To kill the fireflies I set the GI clamp to 10:

Ok, things have improved a little bit, but it's not too great yet either, i.e. the only possibility that is left is to bring out the big hammer out and throw more samples at the problem, but use adaptive sampling to make sure that render times are still manageable. This is the scene rendered with 100K samples and adaptive sampling (0.03 noise threshold, 1000 min samples, 2x2 grouping, expected exposure of 1):

Obviously all this is less than ideal and we will try to improve things in that area, but I hope this helps a bit in the near term. I attached the scene for those who are interested.

Octane is still only a path-tracer. Granted, with lots of bells and whistles, but in the core it's still a path-tracer. The underlying problem of all situations that cause excessive noise is that on those cases it is very hard to find those paths that have a large contribution to the pixel brightness. In this particular case the problem is finding those points on the emitter that are close by the point in the environment medium for which Octane calculates the scattered light. I.e. only a small fraction of the emitter samples have a large contribution, while a very large fraction of the samples have a small contribution resulting in noise.

We will try to improve the sampling of the light sources, but it's not an easy problem to solve in all cases.

[Daniel79]

Thanks a lot, my problem is solved when i disable cast shadow from emitter..now i can render as i want!

[coilbook]

Thanks you all for your feedback regarding the noise in participating media. While there is no easy fix for it we will be looking into it. There are a few things you can do to at least improve the situation a little bit:

As a first step, so get a good view of the problem, I set the camera response to linear (most response curves increase the contrast and thus make the situation worse), the gamma to 2.2, vignette to 0, max samples to 20000, removed the cube transformation, rotated the distribution texture (so I can see it in the preview) and set the camera to look from the left:

05_cube.png

This is the IES distribution texture you are using:

ies_distribution.png

Obviously, a cube is not a good shape for a directed emitter like this. A better choice would be a hemisphere:

05_hemisphere.png

or even better a simple quad:

05_quad.png

I had to modify the emitter power for the various shapes to keep the overall emitter power in the same ballpark. Since the quad is best I'm using the quad emitter below.

To kill the fireflies I set the GI clamp to 10:

06_quad.png

Ok, things have improved a little bit, but it's not too great yet either, i.e. the only possibility that is left is to bring out the big hammer out and throw more samples at the problem, but use adaptive sampling to make sure that render times are still manageable. This is the scene rendered with 100K samples and adaptive sampling (0.03 noise threshold, 1000 min samples, 2x2 grouping, expected exposure of 1):

07_quad.png

Obviously all this is less than ideal and we will try to improve things in that area, but I hope this helps a bit in the near term. I attached the scene for those who are interested.

Octane is still only a path-tracer. Granted, with lots of bells and whistles, but in the core it's still a path-tracer. The underlying problem of all situations that cause excessive noise is that on those cases it is very hard to find those paths that have a large contribution to the pixel brightness. In this particular case the problem is finding those points on the emitter that are close by the point in the environment medium for which Octane calculates the scattered light. I.e. only a small fraction of the emitter samples have a large contribution, while a very large fraction of the samples have a small contribution resulting in noise.

We will try to improve the sampling of the light sources, but it's not an easy problem to solve in all cases.

HI Abstrax,

thank you for explaining. I did not know path tracing works so differntly. Hopefully future AI denoising and improvements will help. Also medium step length should be added like VDBs have. Maybe it can help.

Also Linear response never has a good picture unlike other renderers. Colors are dim, flat and unappealing. Not sure why. Gama is set right.

[PolderAnimation]

I thought I posted this, this morning but I guesse I did not push on submit...

Hi, abstrax,

A while back I had a question about alembic files and material maps. Unfortunately it was not very easy to make those 2 nodes work together. But is it possible to give 1 input pin (instead of all the inputs the abc has) if an material map is connected to an abc file? This would already be a huge help. The same goes for group nodes that are connected to an material map node.

And then an other small question could you ad an additional info pass. I would really like a falloff pass like the node in octane.

Cheers,
Sander

[Jolbertoquini]

Hi Guys,

I have some problems with hairs I use Maya Xgen to generate the curves the alembic option for thickness doesn't work when I change the scale and if I do a manually scale I can see the hair is really strange because at Maya they all have the same scale I don't know if Xgen do something strange.

I realise if I set the scale on decimeters or meter I have to scale to the scene and then I can get the hair but If I want change the thickness I have any control is look like is broked.

Please any help here the scene and the abc cache of the hair

[calus]

Hi Guys,

I have some problems with hairs I use Maya Xgen to generate the curves the alembic option for thickness doesn't work when I change the scale and if I do a manually scale I can see the hair is really strange because at Maya they all have the same scale I don't know if Xgen do something strange.

I realise if I set the scale on decimeters or meter I have to scale to the scene and then I can get the hair but If I want change the thickness I have any control is look like is broked.

Please any help here the scene and the abc cache of the hair

I guess the issue here must be that Maya Xgen interactiv cache generate ABC hairs with a zero Witdh by default or something like that making the ABC file not proper to be used outside of Maya.
But if you enable the "write width" option in the export cache dialog then it works perfectly !
Even Hair UV are correct (sampled from the underlying mesh).

[Jolbertoquini]

Hi Guys,

I have some problems with hairs I use Maya Xgen to generate the curves the alembic option for thickness doesn't work when I change the scale and if I do a manually scale I can see the hair is really strange because at Maya they all have the same scale I don't know if Xgen do something strange.

I realise if I set the scale on decimeters or meter I have to scale to the scene and then I can get the hair but If I want change the thickness I have any control is look like is broked.

Please any help here the scene and the abc cache of the hair

I guess the issue here must be that Maya Xgen interactiv cache generate ABC hairs with a zero Witdh by default or something like that making the ABC file not proper to be used outside of Maya.
But if you enable the "write width" option in the export cache dialog then it works perfectly !
Even Hair UV are correct (sampled from the underlying mesh).

Yes awesome I didn't realize this option "write final width" nice for spot pascal thanks!

maybe this will help the dev to finally support Xgen hairs