Basic Comping of a Vray render

[Gijs]

There are very minor differences i.e. reflection is a bit stronger and not antialiased as well, the color on the blue box is different some what. But over all images are identical.

Has anyone an idea how to deal with the AA problem mentioned here? As that would be the most problematic thing when using this method for stills/print.
Maybe it is no problem for film?

[cpnichols]

since the openexr image format seems to work well in nuke, i assume it's rather an issue how combustion deals with openexr.
maybe some LUT's are required.
any insight from experienced combustion users would be great.

-oconv

I have not tried with Shake either. That would be another option.

[trixian]

There are very minor differences i.e. reflection is a bit stronger and not antialiased as well, the color on the blue box is different some what. But over all images are identical.

Has anyone an idea how to deal with the AA problem mentioned here? As that would be the most problematic thing when using this method for stills/print.
Maybe it is no problem for film?

I seem to remember Vlado making some sort of example with the ugly max dragon, using adaptive qmc AA with the settings 1 / 8. That seemed to help with aliasing in overbright places.

[Morbid Angel]

imo the aa of reflection is something I need to look at closly, and possible vlado to revise this area.

[Gijs]

I seem to remember Vlado making some sort of example with the ugly max dragon, using adaptive qmc AA with the settings 1 / 8. That seemed to help with aliasing in overbright places.

I just did some quick tests, but can't notice a big difference increasing the AA settings.
I think that images will always have these sort of problems when you don't clamp the colors, OR if you don't have a proper tool that does the right tonemapping to capture the full dynamic range of the image.

[vlado]

The AA issue on the reflections is because the samples in the channels are not clamped for AA, but only for the final channel image, whereas for the RGB channel, the "clamp output" option in the Color mapping rollout clamps the colors before the AA.

Newer builds have the option for switching color mapping on and off for individual elements, so this problem can be greatly reduced.

In general, this is a problem that doesn't have a good solution... clamping the colors before the AA produces nicer images, but loses HDR information, and not clamping at all or clamping for the final image only produces jaggy artifacts in bright areas.

Best regards,
Vlado

[Morbid Angel]

vlado, having read your above statement, you would say that any gbuffer reflection pass would be unusable in animation? as there is no way to properly antialiase the image? or does that only relate to high range images?

[dlparisi]

I've finally gotten it to work correctly using the passes but only in Photoshop with 16bit files (my trial of combustion expired). I'm not sure exactly what the problem was, I tried so many things and changed so many setting I'm unclear exactly what I changed. The linear mapping 1/1 helped and seemed to solve most of the problems though. This still may be a combustion problem though.

I did notice one new "problem" though. In refraction areas (even if IOR is set to one), neither the Diffuse, Raw GI or GI pass contain any information on the transparent areas. For example, on a chain link fence with a bitmap opacity map, the transparent areas render out as black for the diffuse and Raw GI pass. To get the comp to look correct you need to add the "Refraction" pass but if you've made any color corrections to the other passes (assuming you did, why else would you go through this trouble) this refract pass will not match up because it is a straight RGB pass of everything seen through the refracted areas.

Is it possible to correct this? Can the transparent areas show the GI behind it? Am I missing something?

[vlado]

vlado, having read your above statement, you would say that any gbuffer reflection pass would be unusable in animation? as there is no way to properly antialiase the image? or does that only relate to high range images?

This relates high-dynamic range images only; also, if you use some non-linear tone mapping as a final stage, the AA artifacts can be reduced a lot.

Best regards,
Vlado

[Gijs]

I am wondering Chris, how do you deal with shellac materials, or don't you use them? because a shellac build from a diffuse and a black shellac with fresnel reflection comes out black in the diffuse layer

[cpnichols]

I am wondering Chris, how do you deal with shellac materials, or don't you use them? because a shellac build from a diffuse and a black shellac with fresnel reflection comes out black in the diffuse layer

Don't use shellac that much... but it does come in useful. I will have to look into that,

[studioDIM]

since the openexr image format seems to work well in nuke, i assume it's rather an issue how combustion deals with openexr.

You're right there: Combustion does clamp EXRs to 1.000 max brightness (the reason why on the one job for which i used them , the comper had a shake and a mac...), so no use trying to get proper Float values from it.

The problem, though, rears its ugly head also with RPFs, IF the Vray framebuffer is used to save separate passes.

A workaround (if sometimes a limiting one) is to save the files as RPF (32bpc, of course) with the "Non Clamped Colour" channel embedded.
Then, in combustion, use a "Show G-Buffer" operator on the pass which has values higher than 1.000 (a reflection pass is a likely candidate), and work it from there.

Also, with the same setup (saving HDR images from the vray framebuffer), combustion screws up the guessing of Alphas (sometimes even inverting it by itself).

My guess is that the problem is with the way Vray's writes the RPFs, as all issues (clamping and alpha channel interpretation) disappear if i save an RPF from the max image save dialog.

I'd love to hear Vlado's take on this, though...

Lele

[tom schuelke]

Hi ,

i thought of making an own thread, but it fits also well in this thread, so ill post my immages here too.

Well what i would like to do is compositing in Photoshop CS, (not CS2). I did follow this thread, and wanted to render out my immages and compose them in Photoshop, cause we often do very big single renders, not so often animations.

i run into problems only with the raw light pass. i just cant find a transfer mode to compose it well to fit the original immage. So now here are my screenshots out of PS, and also the original PS-file.

maybe you see what i am doing wrong..


sorry for the big immages, but otherwise you wouldnt have been able to see enough...


Tom





this is the original immage. i switched on on top of the image stack.






this is the image using raw light with the transfermode "linear abwedeln", i am not quite sure about the englisch word... maybe linear dodge.. ?






the immage with out the raw light pass, which shows, it fits well unto this point to the original image.


at last now the original psd file.


http://www.tom-schuelke.com/forum/ps/psdmanager.psd

[schmyvol_eye]

Im having the exact same problem Tom. and I dunno how to solve it.
Thats what I was getting it in this thread:
http://www.chaosgroup.com/forum/phpB...=116747#116747

Anyone got an Idea how to do this in CS???

BTW, slick looking Windows desktop you got there Tom, how did you get the slick startbutton?

[tom schuelke]

He he

hi schmyvol_eye, here´s the link for the longhorn .... look. i did the screenshots from my partners computer.

but your right. it looks good.

http://www.softpedia.com/get/System/...sforming.shtml

tom