All language subtitles for nuk231-class04

hello and welcome back to nuke 2 3 1

nuke compositing best practices

this is Victor Perez

and in this class we are going to continue the subject

we already started in the previous class

which is optimization for the guarflow

let's move to structuring your script

because this is like secrets okay

secrets are actually

processes that are being over our boat

and they are using resistance of electricity and

and stuff in here

we are processing values of an image

but we can have the same structure of a secret

so we can structure my script in a parallel fashion

or in a serious fashion

the structure of the note based compositor like milk

is because you can just get

everything in separate processes okay

and group them together

a serious and a parallel

but at the end of the day

you can decide

but in a lighter fashion

like Photoshop

for instance

you are getting just things over other things

over other things

and that is actually creating a massive amount of

processing that is not necessary for setting areas

so in here for instance

let me show you what I'm doing

this is a random shot

I'm using um

in this shot

what I want to do is to apply this CG

and then cast the shadow

that is coming from one of the layers

so layer contact sheet again

and here you are

so I have my mask for the shadows

so what I want to do is to apply that mask here

okay promote

apply that color and then apply that with this rotor

that is going to stencil the area of both my monster

the CG and the shadow

so here what I did

is apply the roto to the shadow by using an stencil

and then the same roto to the monster okay

so that my friends

is wrong why

because is exactly the same principle of the masking

when I'm using the same mask to two processes

remember the difference between this and this

so you are having a very similar effect with this

okay you are applying

a primal duplication of an object twice

so in here and in here

okay and that's not very good

I mean the difference is tiny

but you know

when we're talking about compositing

we are talking about detail

so everything is going to be very very tiny

but you know

the detail of everything together

is going to create the photorealistic result

if you are accumulating errors

through the whole compositing

at the end you cannot say what's wrong

because probably is not going to be that evident

but it's not gonna work

because the brain is going to understand

all those tiny issues

okay it's not making sense

so in here the same mat for two applications is wrong

so this is the parallel method I was talking about

so in here I have a process

in here I have another process

so I'm adding a layer over

and then another layer over

so this is kind of a layering effect

so 1 and 2 while in here

what I'm doing is I'm creating

of course a parallel model

one from the shadow

one for the CG

but then I'm serializing them and grouping them as CG

okay so in here

I'm applying my shadow under the CG okay

why I'm using the under instead the over

is because I want my CG to be the main input

remember that the B input is the main one

so if I want to reduce the amount of shadow

or something

I can apply the mix

and the input that is going to remain

is always the B input

okay so that is now getting a parallel

another bottle

and then in here

every time we are using a merge

we are serializing then

so we are serializing before the grand mix

instead of using single pieces

and then assembling

then at the very end

what I'm doing here is I'm putting everything together

and then mixing in a grand mix

okay so here

and applying the same mat

through both elements okay

and the effect of this is the same

as using the chain of effects and the masking

okay so in here

I can say that

everything coming from this input is a CG

everything coming from that input is a live action

so this is the right way

and let me show you

what's difference between this and this

here you are

see this slightly brighter edge

again in this case

you're having this issue

but under other circumstances

with more contrast

that is going to be a real

real issue okay

so this is going to allow you also

to work better with your effects

and this is going to allow you to have more control

through the whole process

and think about

a real fashion of observing the elements

when you are looking at an element

that is cast in a shadow

and then you put something in front of that element

you are actually occluding

the area of both the shadow

and the element together

not separately

okay because everything after that element

that is occluding

is just covering the whole thing

so in a physical manner

this is the right way of applying the effect okay

so this is again

using the parallel mixing

where is more like a layering effect

and this is a serializing of the

script okay

of course in both

they are both the

parallel answers

that in here

the predominant

is the serializing

so grouping

and in here

is the parallel

so just to have a graphical representation

of what I'm doing

in a parallel process

you have the main input okay

and then you are adding

in this fashion okay

while in a serialize

you have this

and this

and this okay

so when you are doing that in this way

you are actually creating that tree shape script

okay because everything is coming that way

instead of having gas

parallel lines coming through the main input okay

and remember that I mentioned before

that we have also implicit and explicit methods

not only for masking a certain effect

but also to create the beauty pass okay

remember that the beauty is the sun

of all the light passes from a CG render

okay so in here I have

at the very end

my beauty pass okay

and this is my beauty

and remember that in here I have many

many components of light

that are being used to create this particular beauty

so let's have a look at this script

and now you can start seeing how massive can be this

so what I'm doing is

in this case

I'm using a file for each frame of each pass okay

in this exercise

I'm just providing you with one frame

because he's not changing when you are using animation

okay so what I did in here

is to apply all the light passes together using

remember the blast operation

all light components are always in blast

never use other operations

because I've seen things like using

for instance

the screen or

and let me tell you

is just playing wrong

okay because to have a photorealistic result

and a physically plausible effect

you need to correspond each light in nuke

with a blast operation

okay so in here

I have my diffuse

in direct diffuse

let me tell you that

all these passes have been generated by Arnold

okay so the naming conventions

and the way of calculating the components

are the Arnold way

um in this image

in this particular example

the image is not promoted by

so is a full image

so there is no black area

in the alpha

everything is white

but in here

I place a and I'm promote

just to get the same effect

when you are working with elements that are promoted by

okay but this is taking us to

one of the fundamentals of the composite

in the beauty

which is every time you are adding a pass

to the mainstream of the beauty

what you are actually doing

is a selective color correction

so you are adding certain values

to certain areas of the image

so that adding operation

is actually kind of a color correction

well in this case

if you are not modifying anything

using the amp remote

is not necessary

because the image itself has been designed to be art

by using also

the anti alias and everything

but not always is like that

and also this method of creating the beauty is

because we are assuming you want to change those passes

the original look I mean

so when you are applying a grade

every time you are modifying a primultivite element

you need to

um pre mold

and then pre mold

the color operation

so that can be quite long

when you have

all these amount of layers

as you have in here

so something that I will suggest you

is to apply the am

pre multiplication

and then having all your passes

am pre multiply

pre multiplying

then together at the very end

so this is going to preserve your edges

and all the semi transparency you have in here

okay so it's something that you should apply as default

the problem is because of course

there is always an issue for everything in here

as you can see

this file has no Wi-Fi

because it's going to be a waste of a space just to use

like in here

maybe 15 alpha channels

that are not different

so they are all the same

so usually software like Arnold is just rendering

and a special pass just for the opacity

which is the alpha

okay in this case

remember that is white

because the image itself is white

what I need to do for every image

if I want to apply a color correction

to a pre multiply image

is just to get that opacity or that matte

and then copy to the stream of my component of light

and then I'm pre multiply

and then placing all together

and then copying the alpha to the final stream

in case you are deprecating the original alpha

and then you pre multiply everything

so that is quite long

of course you can create a template for that

but still it's kind of a massive process for everything

just to have a big script

so this is the natural fashion

of generating the additive synthesis of the beauty

okay by using the explicit way

so if I want to modify one channel I can hear

see I modifying just the emission pass okay

so I can take this particular component

change and adding to the mainstream okay

so when I'm adding my light pass

to create the final beauty

that is a narrative synthesis of the beauty okay

and as you can see

this is exactly the same result okay

let me tell you another way of doing the same thing

but now in an implicit way

so you can judge by yourself what's better for you

and depending on the situation

again you are going to decide

to use the explicit or the implicit

in my personal experience

I find the implicit to work quite better

because you have less notes

and everything is coming together

in areas that you don't need to change

those blast operations that you have in here

I'm not going to change

of course you can just tell me

yeah but maybe I want to reduce a certain component

by using the mix well

that is exactly the same as placing in here a great

and then multiply okay

because that is the fashion of the light to decrease

so the intensity of light is measure in multiplication

and the addition of components

is by using the plus operation

so let me show you

for instance

with this reflection okay

so if I want to reduce my reflection component

I can hear I'm going to place this

here you are

okay so I'm reducing the component

without touching any merge operation in here okay

and that is going to be easier

because if you're using the explicit way

just be all the way explicit

so you can understand that you are modifying 13 nodes

by placing a certain pass

by placing a node okay

so in here I have exactly the same setup

same same setup

so that's the beauty

and here you are

even with the same color correction for the emission

the difference is in here

I have all my passes together in this chain of merges

and then I'm still using just the channels I need

okay because the nice thing here is

I'm not over processing my layers

because here you are

I'm just using RGB and a

okay while in my master

like in here

for instance you know

you have all those channels coming through the output

okay um that's nice if you need them

if you don't

that is actually creating more issues than helping

okay because you are over processing every time

you are applying

for instance

a blur and you don't take care of what

a channel you want to blur

by the fault

you are now blurring all those layer sets

so your blur is going to take ages to be rendered

and in this case is just a simple blur

if you're using a note like median or d focus

or many other nodes that are slower to process

maybe your computer is going to crash

just because you are

using more channels than you actually need

so something that is good is

when you are having a multipass option

is just to remove any channels that you don't need

but we're going over that in a few minutes

because right now

what I'm going to do is

I'm going to apply the same setup you have in here

in an implicit way

so this is my master

as you can see

you have all the channels

and then if you want to apply

the amp remote application to all channels

what you do is just to apply the amp remote

okay but instead of using the RGB

you are going to use the all options

okay so right now

by using one note

you are applying that

and promote application to everything

so everything coming here is now in an unpremaled state

so I don't need to even place in here

the primult obligation

and primult obligation of this note

okay because it's already been unpremotified

and I just need to apply the mask

in case I need to apply some um

as you can see

it's telling me this label

that is emission

and this is not something I place

this is automatically telling me what's the layer said

I'm applying my grade here

so the channels that I'm changing

are the emission channels okay

so if you see in here

what I'm actually seeing at the moment

is my beauty

so if I deactivate this note

nothing is changing

because of course

this is not changing the RGBA

which is the set that I'm at the moment

this is changing the emission

so if I go to a mission

see is changing

but at this point

for me is irrelevant because it's not changing anything

so why is this important in here

well it's because I place this great

after I create the beauty build using the merges

so I'm going to deactivate it for a second

and then here what I have is merge

and this merge is actually taking the direct diffuse

and the indirect diffuse

and put it together in the RGBA set

and how is this coming

well it's because I'm using both inputs

the a and the B

from the same stream okay

so I'm using the multi channel features in here okay

and nothing wrong with that

because in here

in the merge

which is using by the way

the plus operation

I'm deciding here

what channels are contributing to that glass operation

and from what layer set

so from the a side

I'm deciding that these might direct diffuse

and I'm using the red

green and blue channels

and from the B side

and using the indirect diffuse red

green and blue

and then I putting them together

using the plus operation

and the result of that is being placed in the RGBA

okay so with that

I can get the operation in an implicit method

so I don't need to split any shuffles for my notes

and I'm applying that to the direct diffuse

I mean direct diffuse and now

every time I'm adding another bass to the stream

I have to use that pass from the a input

and then preserving the RGBA

in the B input and the output

because of course

I want to comp in the RGBA set

so because in here

in my first merch

I have everything together in the RGBA

as you can see in my output

now my main output is the RGBA

so the B should be at the RGBA

and then any channel being add to the stream

are going to be in the a channel

so in here as you can see

I have a label

just telling me what's the layer and adding

and this is smart labeling

and is smart

because I'm not typing the name of the past

what I'm doing is

I'm going to the no tap and here

and just put in this text

which is just text

so there's no expression for this

plus I'm placing a new expression in here

which is inside square brackets

and I'm saying

give me the value of the knob

a channels and case sensitive

so the a must be uppercase

and what's that

what I'm telling the merch

is just to place whatever is writing in here

okay which is the a channels knob

is going to be display here in the value

so I'm just saying okay

this is adding my direct specular

this is adding my indirect specular

primary specular

secondary blah

blah blah blah

blah blah blah

blah blah blah blah

so you have everything in here together

just by using this chain of notes okay

so another thing that is important is

once you have all you need

what you need to do is get rid

of all those channels that you are not going to use

um be careful

because I'm assuming you know

that you are not going to use those channels again

but no worries

because in case you want to use them again

you can always bypass

this operation that I'm doing here

which is the remove

I'm removing all the channels that I don't need

the remove operation is seen here

and what this is doing is actually telling me

what channels I want to keep or to remove

so you have the operation in here

if you remove the

the channels you are selecting in here

are the channels that are going to be gone

and if you select the option keep

is going to Bing everything

but the challenge

you selecting here

I'm using the keep operation

in this instance

because it's easier just to select the challenge

that I want

that are only three

okay why I'm not keeping the alpha

you know sometimes there are passes that

maybe by mistake

are having an alpha

or maybe an operation is manipulating the alpha

in a way you are not expecting

so I consider

is always a good practice

just to bypass the alpha from the regional

even from the um

pre multiply

okay because of course

even the opacity has been now unpromortified

so you don't want to touch the alpha

so I'm getting my alpha from the regional

using the copy note

and remember

the a input is just getting

the channel to the B stream

so in here I just preserve my

RGB no alpha

and then I'm

recovering my alpha from the original input

okay so with that

I'm getting my beauty

built in here

so I don't need to touch

any of those notes

if I want to modify something

I just apply

a grating here

and I select

what's the layer set

so remember

the grades should always be

before the chain of marriages

in here okay

so in case I want to modify the reflection

I'm going to do it in here

so reflections

here you are so exactly the same see

I'm now tricking the intensity of this reflection

and of course

I can change the color

okay of the reflection

and you know

as any other pass

okay so you place your corrections

for any of the passes in here okay um

just judge by yourself

between this and this stream

of course if you use this

you are being synthetic

in here you are being more explicit

so probably from an empty level

this is going to be easier to understand

but if you are just getting advanced

you just want everything to work in lesser space

and just to be easy to handle okay

so this is the implicit and the explicit way

of getting the additive synthesis of the beauty pass

okay then in here

I have a way of getting a multipass file

and I'm using

for this instance

the merge in the copy

so what I did is I'm placing all my passes together

okay and instead of using the copy

okay and using the merge

why because if you use the copy

you can copy certain channels if you want

so you placing here the red

green blue alpha

and then in here you go

and you say

new channel for the red

new channel for the green

is going to be a bit annoying

just do the whole thing

and still you have a note in here

that is adding that copy

and is going to be way easier

so in here the copy

one of the uses

nothing wrong with using the copy notes

but this instance

just to show a case scenario

when the merch in copy can work

and here you can just get the saying

just copy a layer

but be careful

because if you are copying a layer

where you are actually saying

is that you want the same layer from this setup

to be copy to the same layer in here

so what you are assuming is that in here

in this file

you are getting a layer set

that is not the RGBA okay

because in here

I just have the RGBA

so if I copy

for instance

to this image

I copy let me remove this

and I say I want to copy direct diffuse

for instance

you would see that nothing happens

there is no direct diffuse

even if I go in here

and I go to the direct diffuse

there is nothing

why because he's taking the direct diffuse

layer set from here

and placing in the direct diffuse of this layer set

but in here

you don't have any other rather than the RGB

not even the a okay

so in this case

the copy of a layer set is going to be useless

they need to have both the same layer

just to use this feature

so for this instance

is going to be easier just to handle the passes

by using the copy operation of the merch

because with this

what you can say is okay

I'm going to take the RGB of this layer set

and I'm going to place it

in the direct diffuse of this image

and remember

the operation of the copy is just taking a layer

from here I'm placing in another layer in here

so he's just placing information

okay so in here

I got the direct diffuse

okay so I have

by default the beauty

and now here

direct diffuse

in here I'm doing the same with the indirect diffuse

and at the end

what you get

is all channels together in the same stream okay

so probably

let me be honest with you

if you have all your channels separate

maybe putting them all together today

working in an implicit way

is going to be kind of a waste of time

because you are getting from explicit to implicit

every time you are getting your files

in one past one file fashion

you are prepared to walk with the explicit way okay

there is no sense for you to go with the implicit okay

but in case you want to pre render everything

just to get one file

for handling something

that I don't recommend

if the pipeline of the studio

as representing other ways of doing things

but if you are working as a freelance

or by your own

or you want to prepare the

lighting package for the composity

if you are coming from a lighting background

then this way is the way

but let me tell you a recommendation

if you want to use the multipass

then there is an option

you have to use

which is the open

EXR file okay

um you place it

whatever you want

things to keep in consideration

if you are using passes

that are having more than 16 beat

half floating point

okay so if you are using a pass with 32 beat

then the whole thing should be fit to beat

which is kind of a waste of a space honestly

because you are using a super high level of precision

for something that maybe

for most of the passes

is just half

so nothing wrong in terms of quality

is just in terms of storage

so if one of the passes is having 32 bit

and usually

Aobs like the B

for instance

the position pass or the CDF

are using the 32 beat

floating point

that's a type

okay so here you are

going all the way 32

okay and then

if you are storing a multipass

don't forget something very important

which is the channels you are going to write

so in here right now

you're not even writing the alpha

just the RGB

so if you place the layer contact sheet in here

he just telling me that you have the RGB

a so in here

what you want is actually to get everything

so you put all

and now you are getting everything together

okay so this is in case

you want to deliver

your lighting package

as one single file

and they weren't coming

out of the render package together

so everything is in one file

okay so that is for the components

to be in one stream

now let's review

for a second

the explicit way of the adding

remember that

we were talking about this

and see the difference now

every time I'm adding a pass

of course he's adding that component of the light

but to be honest with you

when you are beating the beauty

sometimes you just want to trick

one of the passes

or maybe not all of them

so when you don't

want to trick all of them

and you just want to create the beauty

probably if you are working with a multipass option

that is providing you

with the beauty

and then all passes separate

like in here

and you want to change only one component

let's say the emission again

he's going to be kind of a waste of processing and time

just to have all those passes together

just to modify one

so this is the additive synthesis

but there is another one

that is becoming more and more popular

which is the substructive synthesis

which is this

and have a look at the difference

I'm going to change the mission

and I'm going to change the mission in here

have a look

exactly the same result

but just those notes

for this case scenario

I'm using the explicit way

because I want to display everything

but again this is exactly the same

if you want to create an implicit way okay

so what I'm doing here is

I'm going to start by using the beauty

which is usually in the RGBA set

and then because this is a multipass file

I'm going to extract the emission okay

and see that in here again

I'm getting this emission

and this is not something that is given by the note

by the fault

this is something that I program myself

and it's really

really easy

you go again to the note

and then in here

you need to write

inside square brackets

the word value

just to write on the value of a certain knob

and then the name of the knob itself

you want to deflay in here

the shuffle has this first knob

which is the in okay

and what I'm telling basically to the label

is just to explain here

what's the channel I'm splitting

in here with my shuffle

so this is going to be way easier to visualize

when I'm working with Moonton channels

using a splitting method of the master file

okay so again

another thing that I'm doing

is to place a postage stamp

okay I'm calling the postage stamp by using Alt B okay

and now in here

the operation and using is the reverse of the minos

the reverse of the Minos

I mean is the a input

is the channel that we are subtracting to the B input

okay remember

this is the from

from is exactly the same as minus

when minus is B in there and a in here

so what I'm actually doing

as you can see

the same same results

but I'm using the from

because I want the master file to remain the main input

okay just to use the correct direction of my script

so here what I'm taking is this component

and because I mentioned that the light

in a linear light color space is always being at

so is using a plus

if you want to remove a certain component of light

what you are going to do

is you are going to subtract that component of light

so in here and just removing

the emission

from the beauty

so if I now

at the same component that I removing here

the result is going to be exactly

the same as the original

okay of course

this is kind of useless just to remove something

just to add it at the end

but in the middle

you can modify what you are adding

so what you are doing is isolating this pass

processing that pass

and then adding that pass to the mainstream

so as you can see here

what I'm doing is just applying the color correction

just to this single pass

so pros and cons of both the

additive and the substructive synthesis is

if you are going to work with just one

or just a few of the layers

I will suggest you to use this substructive way

but for each substructive way

you are using those nodes

which is removing

so splitting the component

removing and then adding

and then the process

while if you are planning to change

several layers of your multipass

I will go directly with the additive way

at the end of the day

the result is going to be the same

the only thing that is changing is how many nodes

you are using for the process

of course if you are planning to change everything

and you are using the destructive way

then you are going to end with way more notes

than the additive fashion

okay so it's something to keep in mind when you are

planning your compositing

but still if you are using the substructive way

you can get exactly the same result as the additive

okay again with this

you can get the implicit way

that I leave up to you

both ways the additive and the substructive

are easy to be synthesized in a gizmo or a group

of course all the possibilities for the pipeline

are available in both

so no restrictions for one or the other

and this is all for this week

we're going to continue with this subject

in the next class

so if you have any comments

questions or suggestions

just feel free to post them in the forums

as always this has been Victor Press for FX PhD

thanks for watching