PDA

View Full Version : basic comp script critique


cheerioboy
05-12-2010, 07:25 PM
So I've been using the downtime I have at work to practice compositing my own layers in Nuke. I just wanted to post my current layout and see if anyone had any comments, be there a better method or just to make sure I'm doing it right. Note that those merges on the left side are doubling up the attached read node to help give the layer more density as they're a bit transparent from their alpha.

digging Nuke!

cheers

http://forums.cgsociety.org/attachment.php?attachmentid=153278&stc=1 (http://forums.cgsociety.org/attachment.php?attachmentid=153278&stc=1)

kag122
05-14-2010, 05:26 PM
Generally I would keep everything
B stream on top
A on the left
Mask on the right.

This is how nuke creates the nodes.

thatoneguy
06-04-2010, 07:09 PM
Note that those merges on the left side are doubling up the attached read node to help give the layer more density as they're a bit transparent from their alpha.



Few things:

1) You should be using the same read node then just forked off. Otherwise it has to read the file twice. That's half as fast.
2) If you want to perform the same operation then you can do that in one merge node. A1, A2, B. A2 will be on the left side of the node opposite the mask input after connecting A.
3) The correct workflow would be to go -> Unpremultiply -> Multiply (Alpha) 1.5 -> Premultiply Now you'll get the same effect.

http://i111.photobucket.com/albums/n134/im_thatoneguy/Merge.gif


set cut_paste_input [stack 0]
version 6.0 v6
Constant {
inputs 0
channels {rgba.red rgba.green rgba.blue rgba.alpha}
color {1 1 1 0.5}
name Input
selected true
xpos -260
ypos -159
}
Premult {
name Premult1
selected true
xpos -260
ypos -74
}
Dot {
name Dot4
selected true
xpos -226
ypos -38
}
set N4edcb80 [stack 0]
push 0
push $N4edcb80
Dot {
name Dot2
selected true
xpos -306
ypos -38
}
ColorWheel {
inputs 0
gamma 0.45
name Background
selected true
xpos -485
ypos -132
}
set N9062100 [stack 0]
Dot {
name Dot1
selected true
xpos -359
ypos -40
}
Merge2 {
inputs 3+1
name Merge2
selected true
xpos -393
ypos 12
}
NoOp {
name Solution_Two
tile_color 0xff
note_font "Verdana Bold"
note_font_color 0xff0000ff
selected true
xpos -393
ypos 155
}
push $N4edcb80
Unpremult {
name Unpremult1
selected true
xpos -260
ypos -2
}
Multiply {
channels alpha
value 1.5
name Multiply2
selected true
xpos -260
ypos 24
}
Premult {
name Premult2
selected true
xpos -260
ypos 62
}
Dot {
name Dot3
selected true
xpos -226
ypos 100
}
push $N9062100
Merge2 {
inputs 2
name Merge1
selected true
xpos -485
ypos 96
}
NoOp {
name Solution_One
tile_color 0xff
note_font "Verdana Bold"
note_font_color 0xff0000ff
selected true
xpos -485
ypos 156
}
Viewer {
inputs 2
frame 1
input_process false
name Viewer1
selected true
xpos -447
ypos 194
}

cheerioboy
06-04-2010, 08:02 PM
Thanks Gavin for the mock-up

I went ahead and tried it and it works wonderfully. Its interesting because I never knew or understood how the premultiply node worked. But it gives me exactly the result I want which I had be achieving before through a 'matte' operation in the merge. Now using the premult I'm able to keep the default setting of 'over'

the multiply node on the alpha is the choice I would go with, although I also tried using a grade node which let me control and multiply along with other options to help keep the edges soft (which was desired in my shot)

and the A1, A2 stuff I had just learned about recently, I probably could have condensed all of my layers into a single merge node if I layered them properly in the node. But I don't know if there's an easy way to rearrange them or enough control in general to go that route.

thanks again for the suggestions!

thatoneguy
06-05-2010, 08:05 AM
Also in this case I just realized that the unpremult-premult stage would be unnecessary if the multiply was set to RGBA and not just A.

(FG/A)*((1.5)*A) == FG * 1.5

CGTalk Moderation
06-05-2010, 08:05 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.