If this is your first visit, be sure to
check out the FAQ by clicking the
link above. You may have to register
before you can post: click the register link above to proceed. To start viewing messages,
select the forum that you want to visit from the selection below.
Exciting News: Chaos acquires EvolveLAB = AI-Powered Design.
To learn more, please visit this page!
New! You can now log in to the forums with your chaos.com account as well as your forum account.
Here are my results using the Area AA filter; additionally I've rendered to .png files (don't have Nuke here to work with .exr's). These are the only changes I made to the scene. Also I used Corel PhotoPaint for the compositing. I've added the following elements together:
specular
caustics
background
global illumination
lighting
refraction
reflection
Here is the comp'ed result:
Here is the original render:
Here is the difference of the two:
As you can see the difference is negligible. I don't know what exactly is wrong with the process you are using though. It might have something to do with Nuke and what it does to images, but I wouldn't know for sure (your results seem brighter for some reason, both the composition and the original render, maybe there is some hidden gamma correction happening along the way??)
The brightness difference was easily fixed....didnt notice it before tho. That changed by switching all nodes to sRGB mode. Still black edges tho...and the strory continues :P
i'm running out of ideas. tried to comp png's as i thought it might be some stupidity in the EXR IO...same prob...will try in PS now...wonder why it worked for you vlado...btw. did you save 32bpp pngs or that 48pp ? maybe it's the clamping that makes it work ?
If all else fails, you can use the VRayCompTex to combine the elements in 3dsmax, that is guaranteed to work (see this for some examples: http://www.spot3d.com/vray/help/150R...nim.htm#3dsmax and note that VRayCompTex is now a part of V-Ray so no need to download that).
Multiplying raw channels in compositing and then adding the results will in many cases lead to edge artifacts, as the alpha isn't treated the way it is intended.
There is a very easy and clean way to do the comp, no edge artifacts at all, and you can even grade your comp in any way you want to without edge problems - just render out everything UNpremultilied, dump the alpha / treat al layers as straight while doing mathematical operations (multiply, add) and reaplly the original alpha in the end.
Gives perfect results even with semitransparent layers with a lot of motionblur, no edge issues at all.
Hope this helps,
Michael
Edit: Just reread the post and saw that i was aswering bs, as you mean any edges, not against BG... sorry then, but this tip might be important anyway when dealing with elements, might save a lot of frustration
BTW - a chance to take a look at your file?
but i cant render unpremultiplied renderelements, can i ?? I already tried unpremulting the reflection with the reflection filter and things like that, but that produced even more artefacts....just different :P
I noticed some weird stuff tho, the alpha is pretty weird in the passes. It has values up to 3.000 or more in some channels....shouldnt alpha be restricted to 0-1 or am i wrong there ?
Alpha didnt change a thing...but finally i could narrow it down. it wasnt related only to the reflection channels. The reflection elements cause edges when the raw*filter way is used...but they work basically when using the finished reflection chan. The prob was caused when multiplying both the raw GI and the the Raw Light passes and adding them...after that there's an edge...So the reason your comp works is that you used neither of them. u used the combined lighting and GI passes as well as the finished reflection and refraction passes...That's at least a reason...but sadly that also means that the raw passes i dig so much are practically useless.
i still havent gotten up completely. i was thinking if i use the raw multiply method, and render one of the two elements to be multiplied without filtering would that remove the double premulting maybe ?
anyways if anyone wants to have a play here's the scene :
The prob was caused when multiplying both the raw GI and the the Raw Light passes and adding them...after that there's an edge...So the reason your comp works is that you used neither of them. u used the combined lighting and GI passes as well as the finished reflection and refraction passes...That's at least a reason...but sadly that also means that the raw passes i dig so much are practically useless.
Like I said, multiplying channels in general is likely to lead to issues. The only situation where this will work is if you render at higher resolution without AA, and then downsample the final results.
Hm....then there's a lot of simply useless elements there. And it gives away a LOT of control (beeing able to tweak ref falloffs, beeing able to balance gi and direct etc etc etc). I'll play around with it some more.
Comment