OctaneRender™ Standalone 1.50

[Interfaces]

Excuse me but I have not followed the release test between 1.20-1.50
and I'm lost.

or is the preview configuration, I can not connect a new camera, I understand nothing.

help me please

Version 1.20

Sans-titre-1.jpg

Version 1.50

Sans-titre-2.jpg

Octane team, please make video tutorials with each new version.
THANKS

[abstrax]

I know that there is an issue with compositing as the alpha doesn't can't work properly through the specular.

All I am doing is using an HDRI background to light the scene, and either have it be the background too, or separately composite it on having used the shadow catcher to get a ground shadow. However, if there is any glass in the object that reflects the ground, it reflects the object's diffuse channel, and if there is any glass that you can see the ground object through, same problem. The workaround would see to be fine, if I could figure out how to do the same in LW, but I was more thinking that it is something that could/should be built into the matte object (or at least an option t).

Problem can be seen in this example.

We are limiting the matte effect to the first bounce (starting at the camera) only, since after the first bounce the environment aligned with the camera doesn't make much sense anymore. To get correct reflections / refractions you would then have to camera project the background on some place holder geometry. We could add a kernel option to not restrict the matte effect to the first bounce only. Would that work for you?

[dionysiusmarquis]

I know that there is an issue with compositing as the alpha doesn't can't work properly through the specular.

All I am doing is using an HDRI background to light the scene, and either have it be the background too, or separately composite it on having used the shadow catcher to get a ground shadow. However, if there is any glass in the object that reflects the ground, it reflects the object's diffuse channel, and if there is any glass that you can see the ground object through, same problem. The workaround would see to be fine, if I could figure out how to do the same in LW, but I was more thinking that it is something that could/should be built into the matte object (or at least an option t).

Problem can be seen in this example.

We are limiting the matte effect to the first bounce (starting at the camera) only, since after the first bounce the environment aligned with the camera doesn't make much sense anymore. To get correct reflections / refractions you would then have to camera project the background on some place holder geometry. We could add a kernel option to not restrict the matte effect to the first bounce only. Would that work for you?

I added more self-explaining images to my Workaround-Post. It already fixed the problem by applying the projected color information to the diffuse color of the matte material (no extra geometry needed):
http://render.otoy.com/forum/viewtopic.php?f=9&t=38911

and this is related:
http://render.otoy.com/forum/viewtopic.php?f=30&t=38653

[ROUBAL]

Hi, I encountered again a failure during rendering animation. The loaded scene was big, but I was rendering only one simple geometry node of it (a bird flying in the sky).

The animation stopped in the middle of the animation of 274 frames. This time Octane was not completely frozen, but the render stopped and didn't want to restart. I had to kill the script.

This time, I took a screen capture showing the error message. I kept on it the screen of the ressources monitor which was open.

The weirdest thing is that I have looked at the rendered images : in fact, even if the render bar stopped in the middle, 274 images have been rendered... but all identical to the first one ! So, 274 times the same image !

[Seekerfinder]

No, it's not a driver issue. Definitely either something dodgy with the MTL file or a bug in the OBJ/MTL import.

Marcus,
It seems to be a [edit: INDEX] FLOAT VALUE issue in 1.5. Default float value in 1.2 was a value of 1.3. In 1.5 the float value default is 3.

Incidentally, a float value of 1.000 returns a zero image. 1.001 brings the image back. Is that intended behaviour?

OCTANE RENDER 1.5 - FLOAT VALUE ISSUE.jpg

Seeker

Yes, an index of refraction of <= 1.000 disables the fresnel effect, i.e. the amount of reflection is solely controlled by the specular channel. But this behaviour was introduced before 1.00, so it's nothing new.

-> I still would like to find out what exactly is causing the difference in the import. Could you post the MTL file that you used and maybe a screenshot of the trunk material node in 1.20 and 1.50? Thanks.

Another interesting discovery is that if I save an OCS file in R1.2 with the OBJ imported and reopen in R1.5, the index float value is back to 1.3 - it obviously retains that value in this case. So the question remains: why does R1.5 change that value (compared to R1.2) when importing an OBJ?

I will send you a PM with a smaller OBJ (please respect copyright) which gives the same results.

Below are the screenshots of the trunk material for both cases.

Smells like a bug...?

Best,
Seeker

ACACIA - R1.2.jpg

ACACIA - R1.5.jpg

[Seekerfinder]

Yes, an index of refraction of <= 1.000 disables the fresnel effect, i.e. the amount of reflection is solely controlled by the specular channel.

FYI, the manual states: "Setting a value smaller than 1.0 will disable the fresnel effect", not <=1.

For when you have time to do a bit of housekeeping on the manual....

Seeker

[abstrax]

Another interesting discovery is that if I save an OCS file in R1.2 with the OBJ imported and reopen in R1.5, the index float value is back to 1.3 - it obviously retains that value in this case. So the question remains: why does R1.5 change that value (compared to R1.2) when importing an OBJ?

I will send you a PM with a smaller OBJ (please respect copyright) which gives the same results.

Below are the screenshots of the trunk material for both cases.

Smells like a bug...?

Best,
Seeker

I had a look at your test scene and the problem is, that the MTL specification doesn't really say, what the Ks parameter in an OBJ material means. It just says:

Code: Select all

 Ks r g b
 
 The Ks statement specifies the specular reflectivity using RGB values.
 
 "r g b" are the values for the red, green, and blue components of the 
atmosphere.  The g and b arguments are optional.  If only r is 
specified, then g, and b are assumed to be equal to r.  The r g b values 
are normally in the range of 0.0 to 1.0.  Values outside this range 
increase or decrease the relectivity accordingly.

Your MTL file has a Ks 1, 1, 1 for the trunk and the branches and a Ks 0.831373 0.854902 0.670588 for the leafs. We have some convoluted logic to figure out which IOR to use. Until now we interprete Ks 1, 1, 1 as reflective as a mirror and therefor set the IOR to 1.0.

After some consideration I think it's best to just leave the IOR at the default (1.3) and change the default glossy specular scale from 0.5 to 1.0 in the application preferences. The change will be available in the next release.

[Zay]

After some consideration I think it's best to just leave the IOR at the default (1.3) and change the default glossy specular scale from 0.5 to 1.0 in the application preferences. The change will be available in the next release.

Personaly I set it to 1.0 (IOR) and spec to 0.04 and roug 0.5. Less fireflies in scenes.
But having options in the OBJ settings to set the IOR,Spec and rougness on OBJ import should solved a lot of import problems. At least get the job done faster instead of fixing every single material in the scene.

[Godstrike]

Hi,

Just installed version 1.5 and have been doing some speed tests.

A scene that rendered at 39 Ms/sec on 1.2 and 1.1 now renders at 23 Ms/sec using direct lighting. Using path tracing the 1.2 and 1.1 versions render the scene at 13 Ms/sec, while 1.5 manages approx. 9 Ms/sec. I have loaded a wide range of scenes and 1.5 is consistently slower. I have the latest NVidia drivers installed (although the same results were present on an older driver before I updated). I have a 3GB 580 and a 1.34GB 570.

Anyone else seeing a similar drop in speed?

[abstrax]

After some consideration I think it's best to just leave the IOR at the default (1.3) and change the default glossy specular scale from 0.5 to 1.0 in the application preferences. The change will be available in the next release.

Personaly I set it to 1.0 (IOR) and spec to 0.04 and roug 0.5. Less fireflies in scenes.
But having options in the OBJ settings to set the IOR,Spec and rougness on OBJ import should solved a lot of import problems. At least get the job done faster instead of fixing every single material in the scene.

The change in the next release restores the behaviour of 1.20. It seemed to have worked good enough and you seemed to have favoured the old behaviour, too (see http://render.otoy.com/forum/viewtopic. ... 95#p163695). I hope that is ok.

The MTL files are just not containing enough information to create good materials and we consider it just a starting point, i.e. in most cases you have to touch those materials anyway to make it look good.