I am having a hard time avoiding the white border around my objects in post. I render out a black environment, like suggested, but I still can’t comp in a nice sky. I have dozens of bitmaps, mapped to tree leaves, and I would like to change the blur to something like .001. I would also like to turn off the filtering. Is there a script that can do this across my scene, or do I have to go into every texture. Is this the right direction to go in, or are you guys mapping a sky to a plane?
Thanks!
If you are just doing single frames, you can use the defringe command in Photoshop (under Layer > matting > defringe) in the newer versions, there is also remove white matte and black matte. This gets rid of the white edges at the edge of you layer transparency
hope this helps.
Andy
Andy- Thanks, I remember that now.
How do you address this problem with animations?
Thanks,
Scott
This shouldn’t be happening in the first place. My first instinct would be that your not compositing your passes correctly or you have issues with your textures/alpha channels in the renders themselves. Without seeing the renders it’s hard to say. But if your rendering on black then you should be able to drop your renders on top of any background and not get white edges. To me that screams of incorrect compositing. If you can show examples that might help us figure out whats going on.
Tim J
Along these lines Tim, if you’re breaking things out into separate elements such as diffuse, spec, reflection and so on, how do you avoid the fringing you get in the end? Is it a case of unpremultiplying all of the elements by the alpha before you plus them all over each other and then reapplying the alpha to the final composited result?
That all depends on what your doing to the buffers in the comp. If all of your buffers are being added (plus or linear dodge) then you shouldn’t be getting any fringing. If your multiplying anything together then yes you have to unpremult the buffers, multiply them together and then premult them by the alpha again. It also matters if your compositing with floating point renders.
Tim J
Ah okay - one of the elements is probably getting multiplied alright. In dd what’s your standard set of passes that you output? It must be the raw ref x ref filt that’s adding in the fringe - cheers for the tip!
Thanks for all the info. How would floating point impact the process?
At DD we change what buffers we save out for each show. That said we normally output quite a few of them but our compositors only use the ones they need. We always output the basic ones and most of the raw ones. Usually the raw buffers aren’t used in the comp but the lighters look at them for trouble shooting their scenes. Then it’s just a bunch of multimattes and object ID’s.
Unless your rendering very dark scenes then there is likely to be values that will go above white. If those values above white are clamped (not working with floating point renders) then you will not get correct values in your comp and you will never be able to recreate the beauty render. If your going to composite your renders then your really better off always working in floating point.
Tim J
I getcha - so you put out raw ref, ref filt and also the normal reflection, and end up just using the normal reflection as a plus to avoid the fringes. I’d imagine it’s a bit of a horses for courses thought but would you ever set your materials a tiny bit wrong so that the passes contain a bit more information to work with in 2d afterwards? In terms of the speed of 2d for broad changes would you bias things towards 2d or really try and perfect the 3d render first?
Not necessarily…The compers will use whatever buffers are needed…but most of the time the raw buffers aren’t needed.
If your working in floating point then you don’t need to have your shaders be a little “wrong.” We always try to get things as perfect as possible in the render and just use compositing to enhance things. That said, our compers are amazingly talented and do all kinds of work on our renders to make them that much better.
Tim J
I getcha, was just thinking that with reflections for example that by setting your falloffs to never run to black, you’ll get a render element that has detail in every pixel, albeit very weak, rather than some of it clipping to black and hence you’ll have more data in that pass to play with in 2d after.
That’s true you could do that. If your only going to add a tiny amount into the blacks though you could probably just raise the black point in the comp and nobody would notice a difference.
Tim j