I hope there was a bet involved... :)
If you need further proof, you could use this script:
set cut_paste_input [stack 0]
version 6.2 v4
BackdropNode {
inputs 0
name BackdropNode2
tile_color 0x7171c600
label BG
note_font_size 42
selected true
xpos 2434
ypos 13246
bdheight 156
}
BackdropNode {
inputs 0
name BackdropNode3
tile_color 0x8e8e3800
label "Convert back to linear\n\n(assuming your destination\napp will
convert\nto sRGB when importing)"
note_font_size 22
selected true
xpos 2103
ypos 14009
bdwidth 286
bdheight 218
}
BackdropNode {
inputs 0
name BackdropNode4
tile_color 0x8e8e3800
label Compare
note_font_size 42
selected true
xpos 2613
ypos 13944
bdwidth 360
bdheight 167
}
BackdropNode {
inputs 0
name BackdropNode1
tile_color 0x8e8e3800
label FG
note_font_size 42
selected true
xpos 2191
ypos 13222
bdheight 190
}
add_layer {rgba redguard1.glow}
ColorWheel {
inputs 0
gamma 0.45
name A
selected true
xpos 2201
ypos 13300
}
Blur {
size 100
name Blur46
selected true
xpos 2201
ypos 13374
}
Dot {
name Dot30
selected true
xpos 2235
ypos 13479
}
set N18ce1090 [stack 0]
CheckerBoard2 {
inputs 0
name B
selected true
xpos 2444
ypos 13326
}
set Nd0e8d830 [stack 0]
Dot {
name Dot31
selected true
xpos 2478
ypos 13639
}
set N985ce2a0 [stack 0]
Colorspace {
colorspace_out sRGB
name Colorspace2
selected true
xpos 2444
ypos 13704
}
set Ned256de0 [stack 0]
Merge2 {
inputs 2
operation stencil
name Merge100
label B*(1-a)
selected true
xpos 2444
ypos 13852
}
push $N18ce1090
push $N985ce2a0
Merge2 {
inputs 2
name Merge101
selected true
xpos 2207
ypos 13634
}
Colorspace {
colorspace_out sRGB
name Colorspace1
selected true
xpos 2207
ypos 13698
}
Merge2 {
inputs 2
operation from
name Merge102
selected true
xpos 2207
ypos 13857
}
set Nd30cc020 [stack 0]
Unpremult {
name Unpremult4
selected true
xpos 2207
ypos 14144
}
Colorspace {
colorspace_in sRGB
name Colorspace3
selected true
xpos 2207
ypos 14168
}
Premult {
name Premult6
selected true
xpos 2207
ypos 14196
}
Write {
name Write2
label "write out here\n"
selected true
xpos 2207
ypos 14263
}
push $Nd30cc020
push $Ned256de0
Dot {
name Dot32
selected true
xpos 2657
ypos 13709
}
Merge2 {
inputs 2
name Merge103
label "Comped in sRGB space"
selected true
xpos 2623
ypos 14057
}
push $N18ce1090
Dot {
name Dot33
selected true
xpos 2755
ypos 13479
}
push $Nd0e8d830
Dot {
name Dot34
selected true
xpos 2917
ypos 13354
}
Merge2 {
inputs 2
name Merge104
label "Comped in linear"
selected true
xpos 2883
ypos 14024
}
Colorspace {
colorspace_out sRGB
name Colorspace4
label "Post sRGB conversion"
selected true
xpos 2883
ypos 14066
}
On Tue, Nov 15, 2011 at 12:30 AM, Ron Ganbar <[email protected]> wrote:
> I knew I was right. (You guys just proved an old argument I had with
> someone).
> Oh, the joys of self gratification.
>
>
> Ron Ganbar
> email: [email protected]
> tel: +44 (0)7968 007 309 [UK]
> +972 (0)54 255 9765 [Israel]
> url: http://ronganbar.wordpress.com/
>
>
>
> On 15 November 2011 10:28, Ivan Busquets <[email protected]> wrote:
>
>> Hi Gavin,
>>
>> As you said yourself, the equation cannot be solved UNLESS you know both
>> variables on one of the sides.
>> In other words, you'd need to have the BG image in order to prep a FG
>> image so it can be comped in sRGB space and match the results of a linear
>> comp.
>>
>> So is there no way to output a PSD or PNG or TIFF which will look the
>>> same as my composite in Nuke over a white background?
>>
>>
>> If you need to get the same results on a white background, you could prep
>> your FG element such that:
>>
>> X = ( (FG * alpha + (1 - alpha)) ^ 2.2 - (1 - alpha) / alpha ) ^
>> (1/2.2)
>>
>> Where X is the FG image you'd want to export to be comped on a white BG.
>> But of course, this will only give you a match when comping the FG over a
>> WHITE BG. If the BG changes, then you'd need to prep a different FG to go
>> with it.
>>
>> Hope that helps.
>>
>> Cheers,
>> Ivan
>>
>>
>>
>> On Mon, Nov 14, 2011 at 12:13 PM, Gavin Greenwalt <
>> [email protected]> wrote:
>>
>>> How are Nuke users handling workflows in which they need to deliver
>>> images with alpha that will be composited in sRGB space not linear space?
>>>
>>> Essentially we have a situation where you would need to find equations
>>> for u and v such that (xy + z(1-y))^(1-2.2) = (uv + z^(1-2.2)(1-v)).
>>>
>>> My initial impression is that it's impossible since the simplified
>>> version of this conundrum would be (x+y)^2 = (u+v) which I believe is
>>> mathematically impossible to solve... right? So is there no way to output
>>> a PSD or PNG or TIFF which will look the same as my composite in Nuke over
>>> a white background?
>>>
>>> Thanks,
>>> Gavin
>>>
>>> _______________________________________________
>>> Nuke-users mailing list
>>> [email protected], http://forums.thefoundry.co.uk/
>>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>>
>>
>>
>> _______________________________________________
>> Nuke-users mailing list
>> [email protected], http://forums.thefoundry.co.uk/
>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>
>
>
> _______________________________________________
> Nuke-users mailing list
> [email protected], http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>
_______________________________________________
Nuke-users mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
