All language subtitles for 4. Additive Keying for Fine Details

In this clip we're going to take a look at additive

keying for following details.

So, additive keying workflow is not an actual keyer.

Instead, it's best described as an image blending tool.

Its strength is in retaining the fine details in wispies,

motion blur and transparency.

The benefit to this workflow is that I can retain details

that keyers may not be able to extract.

However, it does come with limitations.

The main limitation is that the backing region should be as even as possible.

So, the pre-processing workflow may help with this.

Now, it isn't mandatory,

so if you haven't done the pre-processing workflow

to even out the backing region,

you can use extra steps in order to really have a great outcome.

Now, this workflow combines two different steps.

We use our pre-multiplied image that we have used in previous clips and

modules and we also use the additive keyer step that helps to fill in any

extra edge details that we may be missing.

So, let's take a look at our script.

So,

the first thing that we're going to do is go through a few steps

just to see the visuals for what we will be doing.

So,

this is our original image and the first thing that we do

is we're going to compare the clean plate and minus it

from the de-noise source plate.

We're going turn down the saturation to approximately 10%,

so it's very low,

it's not completely de-saturated but we're going to bring down that saturation.

The next thing that we're going to do is place this

image over top of our background,

so you can see here that we've retained just details

primarily where the transparency is.

Next,

let's compare what that looks like against the image that

we had previously without these details.

So, here it is with the details added to the background.

Pay special attention to up here and here they are missing,

so let's just zoom in to see that difference,

so this is with those details and without those details,

so it definitely adds to it,

we can dial this information in so that it works for us completely.

So, now let's take a look at our script.

So, if we look inside of our script,

basically all the areas in blue come from our previous clip

and everything again in teal has been added.

Now, they've been just condensed to be a little bit smaller.

We don't have as many little notes in here,

so if you need to just relate back to the last clip,

please do so.

So, let's get started.

So,

here is our pre-processed plate and the first thing that we're

going to do is basically create a clean plate,

so here I've used a constant,

I've taken the average of this entire background and

that's to create this constant, play into a clean plate.

Now, you can create an actual clean plate from this,

it doesn't really matter but this should be close enough that it's

going to pull up pretty good outcome down here.

Now,

we're going to minus and we're going to take this image and

we're going to minus it against the clean plate to get this

image here and de-saturate it,

so we don't want any of that blue that was in it before,

we don't want any of these funny colors.

These colors are not in our actual source plate.

There is some blue,

so get rid of most of the saturation just to retain the wispies.

Now, we can dial this back later,

we can either make it completely de-saturated or add

more information back as needed.

You will notice that this has a green little note here,

so the green notes basically mean change the settings

to work for your particular script.

If it does not have a green little note,

that means that you can probably leave it as is and use it as a template set.

So, these two steps,

they're not going to make sense quite yet but I will

go into them as much as I can.

So, the first branch that we have off of here is going to be to clamp the black,

so this is breaking our rules,

generally we don't want to clamp any colors in our RGB but for

this workflow with operators it will work.

So, we're going to clamp the blacks, so we're going to clamp them at zero,

so no negative values and we're going to make the super whites at least 1,

000.

So, the typical super white is probably just above one,

it might be one, five, even 15, so 1, 000 is a really,

really safe number.

The next branch that we have down here is to clamp the white values,

so generally speaking we don't want to clamp them at one but it's going

to work again with the process that we have set up,

so don't worry about that, we've captured them up here.

So,

in this clamp we're going to allow for negative values

and that's going to be negative 1, 000,

again more than enough for negative values and clamp the white at one.

So, before we get into how to apply these, let's just shelve these for a minute.

We're going to go up to preparing the background image.

So, we have a checkerboard here and we're going to work with HSL color space,

so that is hue, saturation and lightness.

We're going to go from linear to HSL.

So, what exactly is lightness?

Well,

it's definitely not the same as luminance and it can

often be confused for one or the other.

Lightness is calculated by taking the average between the minimum and

maximum values of the RGB colored image versus luminance which is the

perceptual gray representation of a color.

Each channel represents a certain percentage of the luminosity.

And once combined, they create the overall luminosity of an image,

so for example, red is approximately 21%,

green approximately 72%,

blue approximately seven percent luminosity and when combined

it is the overall luminosity of that image.

So, now that we are working with the lightness,

the lightness is going to be what helps us complete these steps.

We're going to shuffle it into the blue channel,

so we want to make sure that the lightness is into the RGBA,

so we've shuffled the lightness into RGBA,

it's represented in the blue channel here and now we're going

to combine it with these clamped images,

so take the clamped black and multiply it against a branch of what we've just

completed and take the clamp white and also multiply it,

so not the same branch but separate branches.

We will bring them together in a moment.

So, now let's just make some room here to figure out what we are doing.

So again, green,

that means that we are going to dial this back and

forth as we need for our final output.

So, here this is where we're going to adjust the whites,

or the lights rather,

so here is this image here of our top and we're going to dial this,

we're going to dial it once it's over top of the background,

so we're going to dial the gain in here and this one here is

to dial again but is for the dark values,

so let's just take a look at this here very quickly.

It'll make more sense when it's over top the final image,

so here it brings it down and up and again,

we are punching that image from the background through our edges

to make it better integrated and here we go,

here is for the darks and again, with the gain,

lighter and darker.

So, let's take a look what that looks like when they are combined,

plus them together and then let's take a look at what

they look like over top of the flight, so there we go,

there's this information re-added and our final output

which is down here and we've added some details in here

which were not available before, so let's just turn that off,

disable.

I've added some nice information.

So, now what do these two things mean up here?

So, let's see how they actually interact with our final output,

so the gain is going to brighten up and make it more visible for those details.

Be careful in here, it's starting to look a little bit funny,

so don't push it too far.

Just enough to start dialing in that information and the darks over here,

again, so, makes it brighter or darker, so dial both as need be,

again just to make it look like it's well integrated.

There we go.

So,

it's allowing us to work with the darks and the lights to

integrate them into our final image.

Now, one last step,

we're going to pretend that you did not use the pre-processed plate,

so maybe your backing region is uneven,

so I've added a little bit of a dot here to our image,

so here is our dot,

pretending that's our image that's appearing in our

backing region and we have to get rid of it.

So, what we've done is we've created a mat to give rid of that,

so up to our original image, here's our temp key.

We're going to use the same technique as in previous clips,

so go into our alpha,

we're going to blur our alpha just to increase some of this gray in our alpha.

We're going to use the color lookup again to get rid of those gray values,

so into our alpha and just pull it in, so not too far but far enough.

We are then going to dilate this out,

so we dilate this out so that it reaches the edges of our wispies,

this is pretty extreme.

We have some pretty extreme wispies out here and then we blur it

just a little bit so that if there is any information in the

backing region that we don't want, that it nicely fades off into nothing,

so you can always dial this up to ease that transition,

back to full color and then this will be used,

let's just make some room in here, as a mask,

so here was it before, here's it after,

it's masked it out and now we can take a look at our final outputs,

so let's just turn on the switch there.

There we go, now it's gone and onto our final comp.

So, there we go,

there's a really great overview of how we can add even

more extra information to our wispies.

It does have some limitations,

be careful of this edge up in here but overall it's looking much,

much better than what we had coming into this particular script and

it helps us add that detail that we were missing,

so that is with the additive keyer.

Again,

the additive keying process is not an actual key but rather a

blending tool to add that extra information.