If you have an animation where you want to use BOTH the Vector Blur to simulate motion blur, as well as a Defocus Node to simulate depth of field, what is the correct sequence of them in the Node Editor?
Is there a standard, safe way to achieve this?
Thanks as always.
I think that they are mutualy exclusive. The Defocus blur needs z mapped geometry and the vector blur needs geometry but wont respect the DOF edges (circle of confusion), so the smeary effect wont happen on those parts.
Use Cycles and the DOF and vector should be built right but multi frame motion blur woul;d be the most realistic (oversampled frames0 but the most render intensive.
I guess it depends on how exagerated the motion blur is?
Hmmm. This makes sense, unfortunately. The problem is is that depth of field and motion blur are NOT mutually exclusive in the real world, obviously. Thus, it’s unrealistic that it is either/or here in Blender.
However, my need for motion blur was very small. (I just like to have a smidgen of it so that it doesn’t look jittery.) So, I put a tiny Vector Blur on before the Defocus Node for my depth of field. Just a tiny motion blur.
I’m thinking that perhaps if the Vector Blur (for motion blur) could be limited just to the edges of the frame - this could probably be done somewhat easily with Nodes 0 then that might be a decent work-around. If you could just matte out the center when applying the Vector Blur.
I’d appreciate your thoughts…but what are you here for? You’re supposed to be making VSE tutorials, 3PE! (Just kidding.)
Haha, half my tutorials come from people asking questions here! I believe that camera blur could be classified differently to object blur. That is you could just slap on overall image bluring around the edges based on the camera direction, perhaps you could drive a direction blur value from the camera dynamics?
How about using a different render layer to isolate vector blur and fake DoF in post?
I’ve watched a videotut lately where this was demonstrated but I don’t really remember which one.
Yeah, I think that protecting the center of the image, and having the motion blur gradually take effect the further you are from the center of the image, seems like a good way to go. I’ll go the Blender Freenode site and get help in trying to set up some Matte nodes for this (I’m sure that anybody who’s good with them at all could figure it out very easily).
lol
This just made my day…
While there are no guarantees that it could work, it might be worth checking out.
I think that in addition to putting the object that moves on its own render layer, you need to separate your objects into background, middleground and foreground layers. If say motion blur takes place in the foreground or the middle ground you can combine these layers first with the motion blur render layer and either fake DoF or use actual Z values defined by the layer positions.
Haven’t done this myself but I do recall I have seen such an approach demonstrated, just don’t really remember where! :0
Yeah, this is a little too much for me - even if I NEEDED to do so, I don’t think I could go through all of that. But I really don’t expect I’ll ever need to. (And thanks for sharing your laugh!)
I put together this VERY simple little set up to see if I could have the Vector Blur just effect the outside part of the image. Unfortunately, I don’t think this would work.
Problem is that z depth does not take vector blur into account. Perhaps you could generate a fake z from a mist pass that is vector blurred?
Well, I certainly don’t have the ability to do any of the exotic techniques you described there.
But yeah, I think that the fact that z-depth (for depth of field on the Defocus Node) doesn’t calculate the “Speed” for also including a motion blur is the crux of the problem. That’s really what it’s all about, I think.
But as you had said a few posts back, if we could separate the Vector Blur (for motion blur) in a separate pass, then isolate just the edge of the frame (gradually) and overlay that onto the full Defocus Node pass, then that could be a decent work-around. (Not optimal by any means, but maybe suitable until Blender can simulate both motion blur and depth of field together.)
I suspect that, if you do take the time to separate the objects into foreground / middle / background, you will most likely find yourself applying motion-blur selectively only to foreground objects, and defocus nodes (and perhaps other things, like hue & saturation) selectively only to background.
If something needs motion blur, it is “interesting” therefore most-likely foreground. If it’s distant enough that it has been deliberately excluded via depth-of-field, it’s not “interesting” and therefore there’s likely to be no need to worry about motion blur.
