Original subtitles

[Autogenerated] in this clip, we're going to cover channel packing now.

Channel packing is optimization of multi layered e x.

Our files, the exe are files can be very bloated and very inefficient.

And we're trying to make our compositing work flows as

efficient as possible by using a single xar file that

contains all of our render passes.

But at the same time, when we really examine it, it's very bloated.

Now we have 116 megabytes, and right now we have a lot of inefficiency.

So let's take a look at what we have here,

and I'm going to make this a bit bigger and we have our RGB a depth copter,

shadow, puddle, pass, beauty.

We have a stray layer that's a rogue of misspelling.

We've got to get rid of that roto body diffused body directly.

Reflection spec.

For now, a O.

U V and to multi mats.

Now what we're looking for is inefficiencies.

What can be combined?

Well, for one, if we look at the Alfa channels, look at that.

That's a duplicate.

That's a dupe.

That's a dupe.

That's a dupe.

That's a dupe.

This is empty.

This is empty.

These air empty.

This has a ground and doesn't even need anything.

That shadow is important right there.

And the rotor blades?

Well, do we need those or can we combine those with that?

So there's a lot of duplication right here, But are there other duplications?

Well, what we want to do is we wanna optimize for example, rgb a of the plate.

Couldn't we also store the shadow of the helicopter in the Alfa Channel of here?

Couldn't we also declare that something like the for Nell pass?

If we look at this, let's zoom in on this.

Well, we have a duplicate of the Alfa in another quadrant.

Also, if we look at the Red Channel, the Green Jell or the Blue Channel,

are they not identical?

They're identical because they're all black and white passes.

So what we want to do is you want to look at what

are identical in different passes.

So if we look at each one of these, we do know we need RGB and A for the beauty.

And we do know we want rgb a for the helicopter blades.

But we only need rgb of this we need on Leigh read or

a scaler channel of the for Nell, The spec channel.

I I think I see some color in there.

So we do want rgb of that while rgb of the reflection rgb of the body

direct light because there's a warmth to it in a coolness.

So we have to preserve that The ami and inclusion is merely black and white,

so we could just keep the red.

There on this UV pass are G b Will blues empty so

we don't need the blue on multi?

Matt, we've got RGB, but we have no Alfa and we also have rgb over here.

What we want to do is we want to optimize and depth is

Onley in the one Red Channel as it ISS.

So we've got one here and we've got to here.

And if we think about it, we have an opportunity.

Put four channels in one layer one channel set,

and that's the goal of channel packing.

So let's swing over to the node flow and let's put depth

together with UV because there there's red,

which is it's on Leigh read,

because it is the first channel will stored in blue and

we use red and green for the U V.

All right, so here we are.

We have our body UV over here,

and what we'll do is we'll take this one and will take a shuffle copy,

and we're gonna use a shuffle copy on this instead.

So let's just kind of make sure that we're pulling from the right spot.

So we're gonna grab from input.

One are red and green inputs, and you can see there it iss,

and then we're going to grab input to our depth.

So where is our depth?

There's our copter depth.

So let's grab that.

All right.

So you can see that that is coming into our blue on input to there we go.

Awesome.

So we can also flush away our Alfa so we don't have to store those

extra pixels cause we don't need the Alfa for either the u V or

that and weaken store it somewhere else.

So now, after that, we're going to create a new shuffle,

and that shuffles going to shuffle the rgb a into a brand new layer,

and we're gonna call this layer you ve underscore depth and

let's hit rgb a close and then instead of bringing in a

copy of that one and that one.

Let's just remove these over here like that.

We're going to bring in the new copy None.

Put it right there and we're gonna shuffle straight in and copy not the Alphas,

but the layer copy of UV depth.

And now, if we look at the end of our chain, we can see UV depth is R G.

There's our U V S and B is our depth pass Now that just took two

layers removed duplicate Alfa channels and reduce those two

layers in tow one layer without sacrificing any quality or any

usability in the final end.

And we're going to come over here and I'm going to demonstrate this one here is

Let's take the original plate and combine it with shadows.

Let's make those shadows all one.

So let's bring a shuffle Copy over here,

and we're gonna go back to RGB and A and we're gonna input our first

primary input to and put one and we're gonna grab R G and B.

Now the Alfa Channel is gonna come in on input to,

and we're gonna connect input to to the helicopter shadow And what we

should see is if I had a there's that shadow pass.

So I'm gonna leave this one here for now.

We're gonna get rid of this and this and let's there we go.

Let's move this up here and we're make that the B input right there.

So now this is combining rgb and A into that one thing.

Now let's take that a oh puddle and combine it with the other

AO because we haven't another a over here.

So let's remove this from our flow.

We'll bring this over here, so weaken.

Combine these together and let's combine them with a

shuffle copy and we're gonna leave the primary helicopter

input one will put him into the red, will take input to from green,

and then we're going to keep those other two black for right now.

Let's close that.

So we have our red ao.

We have our green ao and we have blue.

We also have an Alfa.

So what else can we store?

What else?

Well,

let's grab some of the multi map because we have 123456

So if we store four of these in one,

then weaken store the additional two in the other.

We're gonna bring these over to those ao areas,

create another shuffle copy, and we're gonna grab this,

like there.

Okay,

We're gonna grab the the blue and the red from this multi map

will leave the green skids for the next one.

The shuffle copyright Here.

We want this to go.

We want to preserve the red in the green we want to bring.

We want to keep from input to red and green.

And we're gonna grab blue from over here,

which is gonna grab the rotor blades,

and then we're gonna put the red into the Alfa.

So we're gonna grab Red of input number one and put

it out into the Alfa like that.

Close that.

So now we should have read, which is our ale green,

which is our puddle ale blue, which is our rotor blades.

And then if I goto a for Alfa, it should be the red things from the multi map.

Great.

I'm gonna go ahead and I'm gonna pack the rest of these images together.

And I will come back with the video to show you how all this

optimization can save some serious disk space.

I have finished this,

and everything is just copying and shuffling into the new area.

And then we render all back out,

and our new optimized version is now 87 megabytes.

Now that's nearly 30 megabytes per frame.

Smaller now may not seem like a lot,

but when you're dealing with image sequences,

every single frame that you go to in nuke is going to

end up creating a input output.

Slow down.

When you have that sort of discrepancy,

we can see that this layered contact sheet we've got

all these different layers over here,

and you can see how fast it refreshes as it tries to recalculate everything.

And if I click over here, it goes a lot faster because we've optimized.

We've gotten rid of a lot of passes,

and we've combined impact the passes together.

If I hit the a key, you can see there's only one Alfa Channel.

There is one Alfa Channel of the body,

one with Roeder's one with rotors by themselves,

and we have shadows.

We could probably skip this one as well and just combine these two to get that.

But I thought we would leave that end in the last clip.

We're going to take that one image file right here to 87 megabyte file,

and we're gonna do a beautiful composite of the helicopter shot.