Original subtitles

Hi, my name is Matt.

And today we'll talk about basic

compositing on the fusion page within Da Vinci Resolve.

This video builds

on what you've already learned in the introduction of fusion video.

So if you haven't watched

that one, I'd recommend starting there and then coming back to this one.

The most basic goal of compositing is to combine

multiple images together to create a final result.

For this first example, we're going to do a split screen composite

where we'll combine the performance of a foreground actor

with the performance of a background actor from a different take.

To do this, we'll use Fusion's tracker

to match, move and stabilize the shots.

We'll use the Polygon tool to draw a mask around our actor,

and we'll use the transform tool to do some fine tune

alignments to those elements.

So let's take a look at our footage

so we can see two

elements here stacked on video one and video two.

And I'm going to go ahead and define a playback region

by hitting AI at the beginning and then O at the end.

And then I'm going to do alt

forward slash to loop my playback,

which is basically the same thing as just hitting play here.

So the first thing you want to do when you're looking at footage

is really analyze what's going on with all the elements

to plan out what you're going to do on the composite.

So for this first element, I can see a take.

One has the performance of the background

character in the passenger seat that we really want.

So if I go and disable that element by selecting it in the timeline and pressing D,

we can now view take two,

which has the performance for the foreground character that we want.

So now we understand what each of the elements are for.

I would say the other thing I notice about these shots

is that there's a bit of camera motion.

So in this shot the camera tilts down slightly and in take one.

If I re-enable that, the camera pans slightly.

And so that tells me that we'll need to stabilize the shot

and match, move them together to get the final result we're looking for.

In addition to that,

I know I'm going to need to draw a polygon around one of our characters.

So if I go ahead and just drop into the annotation tool here on the other page,

I'll probably end up drawing my polygon

somewhere around here to separate

our background

character in the passenger seat from the foreground character.

So now that we've roughly planned this out, let's dive

into fusion and get started on the shot.

So to do that,

I'm going to go ahead and select both of these elements on the edit page

and then right click one of them and select new fusion clip.

And what that'll do is it'll combine both of those clips into your fusion

composition and set that up for the work you're about to do.

So I select that.

And now in the timeline I see Fusion clip two,

and that has also added a fusion clip to into my media pool.

So now if we go to the fusion page,

I can see I've got two media in elements and merge and a media out.

So Resolve has done some of the work for me to set up this composition.

So to just clean up my workspace a little bit.

I'm going to go ahead and select this button here to just single the viewer two.

So we're looking at a larger image.

Next thing I'll do

is I'll look at each of the media ends and determine

which element is which and just do a bit of organization.

So I'll select media in one and I press two to load it

into the viewer, and then I'll play that shot back.

So that shot is the performance we want from the driver.

So I'll go ahead and stop playback and I'm going to rename

this media in one node to Driver.

And I do that by going to the inspector panel.

Right?

Clicking on media in one and selecting rename.

So now

name that no driver so it's easier to keep track of.

Let's view media in two by selecting it

pressing two to load it into the viewer playing it back.

That's our passenger element, so I'll go ahead and rename that.

But instead of right clicking in the inspector tool,

I'll select median two and I'll do a function two.

That's the shortcut key to bring up your rename for that node.

Okay,

now that I have that setup, I'm going to do a little bit organization of my nodes.

I generally like my passenger or my my foreground nodes to be

coming in from the top and my background node to be organized

furthest to the left of the main flow of this tree.

So now that we've set up our shot, we're ready to begin

work and I'm going to go ahead and start with the passenger element.

So I'll select that node, play it back.

So the first thing I want to do is

stabilize this camera motion to just get rid of it.

So I'll go to the first frame here

and I'm going to load a tracker node and I find that in the effects library

so I can show that panel by clicking here

and then I'll twist down the tools menu

and go to the Transform submenu or sorry, the tracking submenu.

And that's where I can see

the camera tracker planner, tracker and Fusion's regular tracker.

The fusion tracker

is kind of a workhorse of fusion, so we're going to use that here.

It's it's very useful in a variety of situations.

So I've dragged that on the pipe that's connecting passenger to merge one

and I'm going to go ahead and hide the effects library for now.

So I want to view this tracker.

So I'm going to select that node and press two to load it into the viewer.

And what I see right away is that I already have this widget

that is in the UI.

So let's go and zoom in a little bit closer and take a look at it.

So this tracker point is made up of two boxes.

We have an inner box here and that box defines

the pattern that the tracker is going to try to find.

And the outer box is the search region

where the tracker will try to find that pattern.

So you can resize either of these boxes by grabbing a corner

or grabbing an edge and adjusting to the size that you want

to move this point, you select this upper left hand corner of the inner

box and you can drag it over a region you want to track.

So what we want to do is we want to track the motion of the car.

Now I'm going to hit command F to go to full screen

and let me play this back real quick.

So what we see here is

the car is moving because of the camera,

but there's also some natural motion within the scene.

This passenger is moving the trees in the background

are moving and even the seatbelt is moving a bit.

We don't want to track any of those regions

because they're not actually being caused by the camera motion.

What we want to track

is the static portions of the scene that show how the camera is moving.

So in this case, that's going to be high contrast parts of the car

like this lock here.

So I'll go ahead and jump to the first frame

and move my tracker down to a

high contrast area like you can see here.

So you want to choose a high contrast pattern like this

because that gives the tracker a more solid pattern to to lock on onto.

If there wasn't much contrast, you may get false tracks or your tracker

floating off the point that you really want to track.

So now that we have that setup, we could track this shot.

But if you only have one tracker, what happens is the tracker

will just be able to solve for any translation movement.

It won't be able to solve for any rotation or scale changes.

So to track for scale

and rotation, you'll need a minimum of three trackers.

So I'll go head back to full screen and I'm going to add

two additional trackers by selecting the add button in the inspector.

So now I have a tracker two and a tracker three,

and I need to find additional high contrast points to track

ideally that are fairly far away from the first tracker.

That way you get a more extreme change for the solve.

So I'll go ahead and choose

this area of the seatbelt

and think I'll choose this

area of the shadow

and I'm going to go and play back

before I track just to make sure I'm not tracking anything like this.

Seatbelts moving on its own

and I'm not tracking that region, so it looks like I'm good.

So I'll jump back to the first frame

and I will select the track forward button.

So the track forward button

will track forward from the first frame no matter where you are.

The button

next to it is the track forward from current time,

and that would be useful if you were tracking from a part

in the middle of the clip.

So in this case, we're starting for the first frame,

so I'm going to go ahead and choose track forward.

Okay, Track is done.

And now you can see if we playback these tracked points track

nicely with the high contrast areas they were supposed to be attached to.

The camera motion is still there.

So to remove that, we need to go to the tracker

and the inspector and from here

we click on the operation

tab where we can apply an operation based off those trackers.

So I'm going to go ahead and drop that down and I want to select Match Move.

And then within Match move, we have various merge opera operations.

Background only

will stabilize the background input.

So foreground over background will actually do a merge operation

with the foreground over the background, adjusting the foreground

motion based off of the tracker that we've already done.

We'll show that a little bit later in this video.

So I'm going to go in select background and then let's play back

to see our result.

So now looks like the car is very steady.

It's not moving anywhere, which is exactly what we're looking for.

The next thing we want to do to prepare this element to be composite

over the top of the driver element is we want to cut out our background

character here.

So what I'm going to do is create a polygon tool, which will allow me

to draw around the region I want to create a map for.

So I'm going to go to the effects library

and I'm going to go to the mask menu.

And here there's various forms of masks we can choose from.

I'm going to go ahead and choose the polygon tool

and drag it into my flow here,

and I'm going to hide my effects library.

So I'm going to zoom out a bit and I'm going to draw a polygon

roughly where I decide to do it earlier.

We'll refine this a bit more later.

And what I'm doing here when I'm clicking

to lay down a point, I am holding that click

and I'm dragging it to actually create a nice curve.

So if I just click and release, I'll create a very sharp angle.

So I generally tend to click and drag to always create smooth edges.

So now that I have that, I'm going

to actually load this polygon into the viewer so I can see how it looks.

And instead of pressing the keyboard to do that, I'm going to go ahead

and grab this node and drag it into the viewer and release

and that'll load that node into the viewer.

So I can see this is where I've defined the Matt.

So White is going to define the region that I want to keep,

and black is defining the region that is going to be removed.

So the last thing I'm going to do is add a little bit of edge softness to this mat.

So I'm just going to drag this a bit.

And now I've got a nice soft edge here

and that's going to help the transition from one of the elements to the other.

So I command have to go back to full screen

and let's view our tracker node

and to actually cut out this character,

I need to connect this polygon to the passenger media in Node.

So to do that, I'm going to go ahead and drag the output of the polygon

and attach it to the mask input of the passenger.

So now we can see we've

isolated that part of the shot.

If we play that back,

seems like we're getting what we want and we'll come back

a little bit later to refine that further.

So now that we have our background character

prepared, we can go ahead and work on our driver.

So to do that, I'm going to go ahead and select my driver

node load into the viewer and watch this element down.

So there's a bit

of camera motion in this shot as well.

Now I need to track that motion, but ultimately

I want to keep this camera motion, but I want to have the other element,

the passenger match, this camera movement.

We'll go to our effects library and go back to our tracking

bin and pull in a tracker and drop it on the flow

between the driver and the merge.

And I want to view that.

So I'll select track or two and press two

and I get my tracker automatically created here.

I'm going to choose very similar points as I did in the first tracker.

Let me jump to the first frame and it looks like

I've lost a bit of light over here, so it's not so high contrast on this lock.

So I'm going to actually track this part of the pillar instead

and that's got a pretty

good pattern to lock on to.

And because I want to track rotation

and scaling as well, I'm going to add two additional trackers

and put them in the same rough positions as I had.

The other.

Okay, now that's ready to track.

I'll go ahead and track forward from the first frame

and that's looking good.

All the points seem to track to the regions that

that I wanted them to

back to full screen.

So to check the track that you've already done

it is useful still to go to the operation mode

and then we can change that to match move and then background only.

That's this is what I used to just make sure my track is looking good.

If my track is looking good, then this car shouldn't be moving at all.

But as I mentioned earlier, I want to keep this camera move,

but I want to apply this camera move to the passenger element.

So to do that, what I can do is I'll change my merge

operation from background only to foreground over background.

And what that'll do is it'll take the tracking information from the driver

element and apply it to the passenger element.

So I'm to before I do that, I'm going to go ahead

and connect my passenger output

and connected to my tracker to the foreground input

and now I can change my merge operation

from background only to foreground over background.

And now I've got the foreground

composite over the background and tracked into place.

So once I play this back

now I've retained the camera motion of the foreground actor,

but the background character is moving in sync.

And just so you can see what it would look like,

let me change this merge operation to foreground only

and you can see that

this passenger now has a camera motion,

but that camera motion matches what was on the driver element.

That can be useful.

But in this case, we're going to go ahead

and just handle the merge right within the tracker note itself.

So I'll change this back to foreground over background.

Now that I've done that, I can actually delete the merge node

because we're actually achieving that merge within the tracker node itself.

So delete that and reorganize things a bit.

And then once we're this place,

let's watch down the shot and see what kind of refinements we have to do.

I can see that this soft edge here is maybe too soft.

And we're also seeing a bit of

the passenger coming through from the driver element.

So let's adjust that a bit by adjusting our polygon.

Now you want to make sure you adjust your polygon on the frame.

You originally drew the polygon.

So in this case, I can see that there is a keyframe marked on frame three.

And I can also look in the inspector

and I can see that there is an animation for the shape

and if I use if I select this

carrot, I'll navigate straight to frame three.

Now, if I modify that polygon, it's going to change that polygon

instead of adding a new keyframe automatically.

So I'm going to go ahead and just zoom in and adjust this polygon

a bit.

Okay.

That seems like that's looking good.

And I'll play back

and it looks like I'm not seeing any of the passenger

or from the foreground element or from the driver element anymore.

So that's looking good.

Let me go ahead and hide this overlay by selecting the polygon.

I play that back.

Yeah, it looks like we're looking pretty good.

If there were any alignment issues between

this foreground and background element, you could refine it a little bit further.

And actually, for example, I'll go ahead and just

adjust this a bit

so we can see if there's any alignment issues.

Yeah, So you can see that

our our door frame here is a bit higher in one element than it is

and the other element.

So what I'm going to go ahead and do while the shot may work

without doing this alignment, I'm going to go out

and do some additional refinement just to make it perfectly aligned.

So to do that,

I'm going to go ahead now to transform Node

on our passenger after this tracker.

Instead of using the effects library.

This time, I'm going to go ahead and bring up a pop up

that allow me to search the tools that are available in fusion.

I do that by heading shift space and that brings up my tool selection.

And there I can enter in transform to bring up a transform tool.

Now when I click add, it will added directly

after the tracker one because tracker one is selected.

So now I've got my

transformed it into the flow there

and from there I can adjust it as I see fit.

Now to do this, I'm going to go ahead and use the split

screen viewer to make this process a little bit easier.

So I'm going to activate the split screen viewer and what that allow me to do

is load two nodes

at the same time in the same viewer and wipe between them.

So I'm going to go ahead and load the driver element

into the side

and I'm going to load the transform

one node into the B side,

and then I can use this

wipe to actually compare the two.

And now we can really see this car isn't perfectly aligned.

So what I'm going to do is I'm going to use the transform tool to

move that element up slightly.

Now we're looking closer

and it does look like the scale is a bit off,

although the alignments looking good.

So I'm going to adjust the scale.

So this part of the window frame matches a bit better

and I'm going to do that in the inspector

by clicking inside

the text box and just dragging it down ever so slightly.

I can use this slider here that would actually do more extreme changes.

If I want to do a very fine tune adjustment, I would click and drag

inside the

the value entry area.

So that's looking a little bit better.

Let me wipe between those two.

That's looking a lot closer

than before.

So I think we're pretty good to go

now to get out of the split screen view.

I go ahead and click this button again to deactivate it.

And now let's view the results in our tracker node.

So I'm going to click that and press to load into the viewer

command F to go full screen and then play.

So now our alignments much better.

I do see that there's a bit

of fringe coming in there, so I'm going to go ahead

and adjust my polygon one more time.

There we go.

Oh, and I added on the wrong frame.

So you can see here I've added this, I've done this adjustment

on frame two, but I also have a key frame on frame three.

So what's going to happen if I go to frame three, it's going to animate.

I don't actually want that to happen.

So I'm going to go back to frame two here and I'm going to remove that key

frame by clicking there.

And now I'm going to go to frame three and do my final adjustment.

And that

should do exactly what I'm looking for.

Okay.

Now are shots looking pretty good?

Let me play this back one more time

because there's one last test that we'll have to do so we can see here

on the bottom edge of the frame, we're getting a bit of a double image.

And what's happening there is our passenger

element is a bit smaller or bit shifted up from our driver element.

So what we need to do

is actually blow up the image slightly just to crop out those edges.

So I'm going to add a transform node after this tracker to Node.

And to do that, I'm going to go ahead this time instead of using the effects

library or the tool search menu, I'm going to go ahead and use our toolbar here

and drag the node directly down into the flow where I want it.

So let's view

that node by dragging it into the viewer and releasing

and now reviewing the transform node.

And I'm going to just scale up the image slightly to crop out.

This double image we're seeing here.

Okay.

And now we'll play back and see if that's giving us

the result we want.

Yep. I can't see any frame edges anywhere.

Shot's looking good.

So we created a fusion composition that brought in two elements.

Our driver and passenger.

We renamed both of those media ends appropriately.

We used our tracker one to stabilize the passenger

and then we added a polygon to the passenger mask

input to cut out just the passenger

and remove the driver

and then we added a tracker where we tracked the drivers motion,

but we used a different mode in the tracker

to actually merge the foreground over the background,

which uses the drivers tracking information to match, move the foreground.

And then finally we added a transform on the passenger side of things

to align her element a bit closer.

And we use a final transform

to crop out any of the edges of frame that we were seeing

from the passenger elements that created our final composition.

So that's how we did our split screen, which is a very common task

in compositing, combining multiple images together to create a final result.

Another very common task in compositing is to replace a sky

that was shot on the day with one that matches other shots more closely,

or one that just looks better

to do this task.

We'll use a little mix here.

We'll use some map manipulation tools like a row dilate

and map control, and we'll have to match move the sky.

But instead of using a tracker node, this time we'll use a tracker

modifier within the merge node to do that operation.

So let's take a look at the footage.

So here we have two elements stacked over each other

and I'll go ahead and set an endpoint and an hour point

just to loop that.

So we've got two characters here and we have a very blown out sky.

It doesn't look very appealing at all.

And then below that element,

if I select the one on track two and had DE to disable it,

I can now see the sky.

So we have a sky element

that we want to use to replace the sky that was originally shot on the day.

And like any other shots, you're starting for the first time.

We really want to look at the footage and analyze it

to determine how we're going to do the work that we're going to do.

So in this case, as I'm looking at this foreground element,

the sky is very, very bright.

So it seems like I'll be able to use a lumix here

to keep out all the brightest parts of the image

to then insert the new sky.

So that's that's one part of the plan.

The other part of the plan is there's some camera motion.

So I know I'm going to need to track this plate

to match, move the sky element to it.

This sky element is just an individual tiff,

so there's no camera motion built in there.

So that's the rough plan for this shot.

So if I disable the foreground

element here on track two, I can now see the sky element.

Now sky looks great,

but there's a bunch of these mountains and buildings that we don't really want.

So what we can do is select that clip

and go into the inspector and scale that up or zoom

to me, pause that

to crop out all the unwanted parts of the image.

So if I go ahead and hit this disclosure in the media pool,

it'll tell me the resolution was resolutions much bigger.

It's 3800 pixels wide or so, whereas

the actual footage itself is HD footage.

So the thing to consider here is that fusion,

if we take these two clips

and like the last shot, we create a new fusion clip

and then go to the fusion page.

We're creating a composition that's the same resolution

as the timeline, which in this case is HD.

That works fine for this foreground element,

but our background element

is actually an HD

when there's a lot more resolution that's available to us.

So there's a slightly different way to to handle things

so that you maintain resolution

of all the source files independently.

The other thing to consider here

too, is when you're creating a new fusion clip from the edit page,

if you do any size changes on the page, it'll crop the image here.

So if we ever wanted to change how far we were

zoomed in on the image, we couldn't do that at this point.

So let's say I added a transform node

and I viewed that

and I wanted to change and zoom out a bit more.

Well, the image has been cropped

and that cropped happen crop happened on the edit page.

So what we're going to do is we're just going to reverse out of this

and undo everything we did.

And instead of actually making a new fusion clip here,

we're going to go ahead and disable the foreground

and then we're going to go directly to the fusion page.

What that'll do is that it will make a fusion composition

that matches the resolution of the clip that's in the timeline.

So before we jump into our shot,

let's talk a little bit about resolution independence.

So in this case, we've got a foreground element and a background element.

And if we look in the media pool where these files came from,

I can click on this all three

dot metadata menu here and it will show me the resolution.

I can see that the foreground clip is HD

and this sky frame is over.

HD It's 3800 pixels or so wide.

So we want to keep resolution independence

and the full resolution of those source files.

Now, if we created this fusion clip like we did on the last shot,

what it will do is it

will actually resize all the elements to the timeline resolution.

So let me do that real quick and show you what that looks like.

So if I go ahead and select both clips

and I say create new fusion clip,

it'll create a fusion clip.

And then I go to the fusion page

and on the upper right hand corner, in my view, or I can see that

I've got an HD image and that's based off of what the timeline resolution is.

And each of these elements, if I view them independently,

well, even this sky element is HD, even though

the original version of that Sky element was about 3800 pixels wide.

So we want to actually retain all that resolution for as long as we can.

So let's go back to the edit page and undo what I just did.

So instead of making a new fusion clip,

you can actually go directly to the fusion page and work that way.

So I'm going to disable the foreground clip and now when I go to the fusion

page, it's going to set the Compositions resolution

based off of the clip that's visible in the timeline,

which in this case is the sky, which is about 3800 pixels wide.

So if I go to the fusion page now, I only have one media in Node,

but my composition

resolution is 3800 pixels wide, which is exactly what I'm looking for.

I want to maintain that resolution of the sky element,

so I'll hit command F to go to full screen.

I still need to get my foreground

element and I can do that by opening up my media pool panel

and then dragging in this element into my flow.

And so now I have a media into if I load that into the viewer,

I can see that it maintains its original resolution of HD.

And so this is how you would make sure that all the elements

that you're bringing in maintain their highest resolution images.

You would actually add them directly from the media pool.

Let's do a little bit of organization in this fusion cup.

So this medium too, I'm going to go ahead and rename that

to foreground

and I'll rename my media

and one after loading into the viewer as

Sky. Now that

I have both these elements, I can go ahead and hide the media panel

and now I'm going to go to full screen command F to go to full screen

and I know I want to compositing my foreground over my sky.

So I'm going to go ahead and add a merge node, and I'm going to do that

by actually dragging the output of my foreground node

and attaching it to the output of my sky node.

And that'll automatically create a merge node for me.

So if I view that, I can now see that my foreground

is smaller in frame than my sky.

And that makes sense because my canvas right now is 3800 pixels wide,

whereas this foreground element is only HD.

Now the one thing is we want the final output

of this composition to be HD because it's going to go back to the timeline

as HD, and we want that to match the foreground resolution.

So the way to make sure that your composition

output matches a particular resolution in this case.

HD Is that the first element in your node tree has to be a HD element.

So to do that, we're going to go ahead and make a background node

to act as our first element in this processing tree.

So to do that, I'm going to go ahead and drag

in a background node from my tool panel here,

and I'm going

to view that currently this background node, it

matches the resolution of the composition, which is 3800 pixels wide.

And if I want to modify this, I can do that

in my inspector by clicking image

and then I can see I have my resolution

here 3800 by 2100 roughly.

If I want to override that, I have to do select auto resolution

and then I can type in the resolution that I want.

You can see now I have a canvas that's 1920 by 1080,

and all I need to do now is add a merge node.

So I'll do that with shift enter to bring up my tool search

type and merge

and I'll connect this background and to the background input of that merge

and this guy to the foreground input of that merge.

Now if I view that merge,

I have the sky inside of an HD canvas.

Now this green border, this green box that's outside of the canvas

actually defines the available image from the sky.

So we're not losing any of the resolution or the image that we had before.

We're maintaining all of it.

It's just not shown in the canvas.

So if there was a transform, we could bring that back into the frame.

So in this case, you can see we have

all that sky available to us.

So I'll just go and undo that

last thing we need to do to prep this.

This node tree is change our background

input from merge one to actually come from merge two.

Now if we view this merge one,

we've got our foreground over our background

so we can't see the sky that we want to replace.

What we need to do is kick out the sky from the foreground

so we can see through it to the sky that's underneath.

So to do that, we're going to go ahead and add a little mark here

and make a little bit of space in my cap.

And I'm going to find the Lumix here under the effects library

tools.

And then, Matt,

you'll see there's a variety of different colors here.

The Lumix here is the one we want.

We'll drop it into the flow.

So the Lumix here is going to try to create an alpha mat

by a particular channel.

In this case

it's going to be luminance, so that'll be the overall brightness of the image.

You can also select an individual channel like blue,

green, red, alpha, etc.

but in this case Luma should give us what we want.

So instead of viewing the rugby image

here, we're going to want to view the Alpha Channel instead.

So I'm going to view only the Alpha Channel that's being created

by selecting this

dropdown and selecting Alpha.

I'm still viewing the merge node actually what I want to view

is that Lumix here, so I'm going to go ahead and select that and press two.

Now you can see that the map we're creating, the alpha we're creating,

is based off the luminance of the image.

I was going to hide my effects library.

I want to adjust this.

So all this foreground stuff is totally black and our sky is totally white.

So do that.

I will drag my low to the right

and then I'll drag the high down a bit as well.

So I'm pretty well isolating the sky now.

We still see a lot of bright spots

in the foreground, part of the image that we don't want to keep out.

So what we can do at this point is actually

adjust this mat

and we're going to do that with a tool called Arrow Dilate.

So we're going to add arrow dilate from the effects library.

So we're going to reveal the effects library

and then arrow dilate is going to be found under filter.

And I'm going to go ahead and drop that in the flow here.

Let's view that node.

And so what Arrow Dilate will do is it will grow or shrink your mat

accordingly.

So if I go to the amount here, I can shrink that.

And what that does is it actually

eliminates a lot of those small holes we were seeing in our mat.

So I pretty much eliminated all of those little holes

in the alpha that that we didn't really want.

What I've done, though, is that's actually eroded this edge.

So we want to restore that edge from where it was.

So I'm going to add another arrow dilate and let me make a little bit more room

in my cup.

So I'm going to add another arrow

dilate after that one and view it.

And so what I want to do now is I want to bring this edge back

and let me go ahead and reset that and zoom in.

So as I drag

this to the right, I'm dilating out that mat now and doing

so is going to bring our edge closer to where it really should be.

So roughly there.

So instead of being my Alpha, let's view the job and we can see here

we're not seeing the foreground at all,

but we have a really nice mat we've prepared.

So what we can do is we can actually combine this mat.

We've created this Alpha with the original ARG image

and then do some interpretation from there.

So to do that, I'm going to go ahead and organize all of these three nodes

over to the right.

These were the nodes I used to generate this man

in the first place, and I'm going to add a node called Mat Control.

And so mat

control is found under the mat bin,

then we can drag that in under the foreground here

and I'm going to feed the foreground into the background input of mat control,

and then I'm going to feed my a-roads pilot output,

which has my, my fully prepared alpha into the foreground input of Mac control.

So I am just attaching that to the green part here, the green input.

And let's view that Mac control.

So now I have my RG B image.

I need to set my combine mode to combine Alpha

and now I've got my RG image and then if I view

only the Alpha channel, I have my mat,

I can still kind of see my foreground here and that's because

I still have the full RG image and I've just combined the alpha

into this node, but I actually haven't multiplied the image.

So multiplication basically means deleting out anything

that's black and retaining anything that's white

and then having a gradation of that

in between for anything that's not perfectly black or white.

So I'm going to go ahead and click post multiply that's giving me just the sky.

I actually want the foreground in this case.

So I'm going to select Invert Mat

and now I have my foreground, so I've got my foreground cut out.

If I zoom in, that's given me roughly what I want, but it's a really hard edge.

So I'm going to go back to my lumix here and I'm going to go ahead and add a blur

to that.

And now I've got a nice soft

edge around the characters, so that's a good place to start.

Let's go ahead and do command F to go full screen and hide our effects

library and let's play that back and see what we're looking at.

So far, so good.

And to double check my mat for any additional holes that may have popped up.

I'm going to go ahead and load my arrow, dilate into the viewer,

and as I play it back, it's looking really good.

But I do have some edges here that, are still poking through,

and that's probably because I added that blur.

So what I'm going to do to fix those remaining holes

is I'm going to actually create a polygon mask to isolate the areas

that I don't want to keep and subtract that out of that that key.

So to do that, I'm going to go to my first frame here.

I'm going to add a polygon tool with shift enter

to bring up my tool search and type in polygon press.

Enter now I've got a polygon tool

and I'll just draw a very rough

polygon here

to define the region

that I don't want to keep.

Okay.

And I'm going to go ahead and add

a little bit of softness to that.

Now, the Polygon node acts like background node

in that it conforms to the resolution of the composition.

So let's go and view that polygon node loaded into the viewer.

And if I zoom out, I can see the resolution of that is 3800 pixels wide.

My image is actually in HD image I'm keying right now,

so I want this to actually conform to an HD frame.

So to do that, I go to my image tab on the Polygon tool

and I

choose my output size instead of default to be custom.

And from there I can disable use

frame format settings and set my width and height to the resolution.

I want.

Okay, so have 1920 by 1080.

That's exactly what I'm looking for.

Let's view our road dilate again

and I want to feed this polygon tool into my little mic here

into the garbage man input, which basically is defining the garbage,

so to speak, that you don't want to be part of that tool.

So instead of dragging the output of the following,

quote, gun tool and attaching it to one of these inputs,

I'm going to go ahead and click and drag this output while holding alt

and then just drag it over this tool and then release my left click mouse button.

And what they'll do

is it'll bring up a context menu that will allow me to select an input.

So in this case, I want to select Garbage Man.

This can

be really useful if a tool has a lot of inputs

and you want to find a particular input, but the colors really are helping you,

such as in this case,

garbage man is one of the gray inputs, so it's a little hard to find.

So the polygon tool is attached to the garbage man, and now we can see

that all of those kind of defocus areas that were there are now gone.

So if I play through this clip

getting a much better result,

I can see a little bit is still peeking in there at the end.

So what I'll do is I'll go back to the frame where I originally

drew this polygon

and I'll just adjust the spleen a bit

to get a little bit closer to our actual edge.

And now let's play back to see what we have

maybe a little bit more.

It seems like there's just a bright spot

that pops up there at the end

playing that back.

Yeah, it looks pretty good.

Okay, so now let's view our Mac control node again.

That's giving us the result.

We want and we want to change the input

to our merge node from the arrow deployed to this mat control.

So I'll go ahead and drag the output of the Mac control

on to the foreground and put out the merge one.

And let's view that merge one.

And now we've got our sky behind our actors.

That's getting us going in the right direction.

It's still not perfect.

The sky looks fairly fake at this point and it's also not moving.

So the sky needs to move in conjunction with the foreground element.

So we're going to do a bit of tracking here.

So let's do a

quick recap on what we've done so far.

I'll go ahead and pause playback here.

We brought in our sky node, followed into the viewer,

which was very high resolution. We did that.

So we were going to be resolution.

We're merging that over a background node

that we set to be HD resolution,

and then we brought in our foreground node, which has our actors.

We did some little mixing work to clear out the sky

and then we did dilation to improve and remove some of the remaining

foreground bright parts, which we did with the A-Road.

And then we did a second arrow to bring our edge back to where it was.

And then we also added a polygon

to as a garbage mat to the lumix here to get all the last bits

that were still left over from the processes before.

So all of

that was to generate the alpha that we then combine

with our original foreground using the Mac control node.

And so we're combining it and the Mac control node

by setting the combine operation to combine Alpha.

And then we have to post multiply that image to remove the parts

of the shot that we don't want based off of the alpha that we generated.

And then we had to actually invert that mat because we had selected the sky

and we wanted to actually keep the foreground and not the sky.

So the invert allowed us to do that.

So instead of using a tracker node itself, we're going to use the tracker

functionality embedded within another tool using a modifier.

So let's get into that.

So I've selected the merge to Node

and what I want to do is actually attach a tracker to this merge center.

So if I right click on the merge center in the viewer

and go down to merge center in the context menu

and then go to modify with tracker position

that will create a tracker within this note that I could

then use to track the foreground and apply that motion to the sky.

So I've created

that and what I see here is a tracker pops up on the screen

and also this modifier in the inspector shows up as well.

So I'll go ahead and click on modifiers

and this looks like just the regular tracker inspector menu

that you would expect to see if you use the node directly.

So I can choose my tracker source instead of the background.

I actually want to track the foreground so I can load it as a tracker source

by dragging that node into the inspector and then releasing.

And now I'll be tracking the foreground.

I also want to view the foreground because that's where I'm going to be tracking.

So I'll go ahead and select foreground, load it into the viewer,

and then click back on my merge to

and go to the modifiers page to start my track.

I want to choose a high contrast area,

so I'll choose this part of the rocks

and adjust my search region and my pattern

and I'm going to go ahead and track forward

from the first frame.

Okay.

Now the track completed.

Let's go ahead and look and see how that track looks

as I play back. It's going very well.

It's saying attached until a little bit further

along here, the actor crosses in front of the tracker.

So that's a pretty common issue that happens,

happens just over the last few frames.

Me jump to the end here and I'm going to step back with the back

bracket or the left bracket button.

We can see the track is great up until about

that frame

and I can see the center is popped off its original location.

So let me go back one frame.

Now, you can fix this by actually

changing the region of where you're going to track.

And to do that, we're going to go to the inspector,

expand the tracker.

And what we need to do is change

the path center to track center, append.

And what that allow us to do is it will allow us to change the region

that's being tracked but still keep the same anchor point.

So I'm going to go ahead and find a different area of the image to track

or just choose this part of the mountain because the actor is not over it

at this point.

And so from here I just need to track these last few frames.

I had issues.

So instead of use the using the track forward button,

which would start from the very beginning, I'm going to use the track forward

from current time button so it will overwrite the bad

keyframes and place new keyframes based off this new region.

So let's click that button.

All right.

So now as we playback, we can see this tracker

sticks really nicely all the way through the clip

and even when the actor moves in front of it, it doesn't

get modified.

Okay, let's go back to full screen with trend F

now that we have our tracker, let's go ahead and view this merge note

and see what it's doing.

So as I'm viewing that I can see now

the sky has inherited the motion of that foreground element.

So just to recap, instead of using a tracker node,

we created a tracker modifier

to modify the merge center by right clicking

right clicking on the merge center and going to modify with

and selecting tracker position and that will embed tracker

that you can then interact with in the modifiers page.

So now that we've applied that tracker, let's view

our last merge in this tree and see what we're looking like.

It like the sky is tracking really well.

It stays attached to the foreground.

It still doesn't look that

perfect.

It kind of

looks like it's pasted in there instead of really blending with the original shot.

So we're going to do some additional work to make this sky

a blend better with the original.

So there are some tricks we're going to do to actually allow

the sky to blend better with the original image.

And so we'll start off by

mixing in a bit of the original sky,

which will help the new sky blend better.

So we'll do that by adding a merge node.

So let me make a little

bit of room in my composition here

and I'm going to add a merge node after this.

Merge two and I'll go ahead and just

find that with my tool search menu

and then I'm going to attach the foreground

and put this to the foreground element directly.

Now let's view that merge three

and we can see that we're not even seeing any of the sky there.

What we want to do is actually bring down the opacity of the foreground a bit

so that we're controlling the mix between the replacement sky and the original sky.

So I'm going to go ahead and select this merge three, go to the blend control

and drag that down a bit.

So something about there.

So right now we've blended in the original sky

with the new sky and that's brightening up the new sky

a bit and making it match a little bit better.

Let's view or merge one node

and we can

see already that it's looking better than it was before.

And just to do a quick comparison on the before and after, I'll select my merge

three node and I'll hit the disable and enable button here

and we can see that's already a more convincing result

for the new sky.

What I do see though, however, as our edges have a bit of a kind

of a halo around them, so that's not quite matching yet.

We can make that match a little bit closer using another merge node.

But let me talk a little bit

about why you would do that and what's the theory behind it.

So I'm going to go ahead and go into the dual viewer mode.

So I'm going to allow I'm going to load the current status

of the background sky into the left viewer.

And actually, I'm going to do a split screen comparison in this left viewer.

So I'm going to click that button

and I'm going to load the foreground into the right side of that viewer.

So as I wait

between these, I can see that the sky

replacement is a bit darker than the original sky

and that these rocky edges

are darker than they are in the sky replacement.

So what I can do is actually create a merge node

that will compare the two images

and keep just the darker pixels between them.

So what that will do is if I keep the darker pixels, that will keep the sky,

and if I keep the darker pixels from the foreground,

that's just going to keep this rocky edge and not the sky.

So let's go and add a node and do that.

So I'm going to select Merge three, I'm going to add a merge node

with my tool search

and I'm going to connect my foreground

to my foreground input of the merge node.

So let me load this merge for

into this right buffer here.

And instead of a normal operation, I want to use the darken operator.

The darken will look at each the background

and the foreground inputs and it will keep the darker of the two pixels.

So I'll select darken and now I've got a nice sky

and I've got my edges, which is exactly what I want.

And in the right viewer I can see merge two

which now doesn't have that halo issue we are seeing earlier.

So we're looking pretty good. I'm pretty darn good.

I'm going to go back to a single viewer by clicking this button

and let's see what our final results look like.

It's looking pretty darn good.

I would say.

I need to maybe blend in a little bit more of the original sky,

so I'm going to select my Merge three node and I'm going to just move

this blend up a bit.

So that's a little bit more convincing for me.

And we retain our nice edges along the mountain side here

and that sky looks like it actually now fits and tracks well into the image

and that's how you do a basic sky replacement in fusion.

So to recap,

we brought in all of our individual elements.

We cleared the foreground,

we did some manipulation to that mat, to compositing it over the sky.

Then we used a tracker modifier on our merge two node

to match, move the sky to the foreground element.

And then we did some additional composite thing with the merge nodes

to blend the sky better with the original sky.

And that's what we achieved through merge three and merge four

to give us final result which we're seeing right here.

So thank you for joining me in compositing.

The split screen shot in the sky replacement.

If you have additional questions, please check out our forum where the developers

and professional users are interacting with each other on a daily basis.

Thanks again.