Hi, I'm Gray Marshall from Gray Matter Post and today we're going to talk about DaVinci
Resolve 9 and the differences between the serial, parallel, and layer nodes as various
approaches to color correcting.
I want to start out by saying that this is my point of view and that there is no right
way nor wrong way in using any of these approaches.
This is where you get to make your own choices and your own looks.
As a quick overview, I'd say that the serial node is the most direct and normal way to
create a color correction, applying one color correction at a time and then adding on the
next color correction or color instruction on top of that.
Sometimes, though, you want to do two things at once, applying two corrections at the
same time.
This is where parallel comes into play and lets you do two or more corrections or instructions
at once.
On the other hand, layer node combines images, not corrections, is much more of a composing
tool, allowing you to combine images in a very specific way.
Sometimes these images have the same mother image, same source, and sometimes they can
be different images, such as a grain image.
Now let's get a little deeper and I'll show you what I mean by all this.
The first, simplest, and most common approach is the serial approach.
This is what everyone uses and it will make up at least 80% of the work you do.
It starts off by taking your first image, then we'll add a serial node, drop the gamma
a little, then go ahead and add another serial, and here we'll drop the saturation quite a
lot.
Add another serial and give it a bit of a pink tint.
So we see that we go one after another after another.
This is the way serial works.
You do one thing, then another thing, then another thing.
One of the things here is that order is important.
We start changing the order like this, swapping the tint with the desaturation.
We see that this creates a different image.
Let's undo that and get back to where we had it before.
The tint is much more apparent in this version, so you see the order does matter in this approach.
Really everyone should be familiar with this.
Some will try to correct with only one or two nodes, others with many more.
Almost every color correction includes some serial work.
It's your choice, it's your approach.
Now let's get more complicated by looking at the layer node first.
Let's say the client asks us to do a full desaturation on the overall image, but keep
the signature red of the motorcycle.
For that, we'll add a serial node.
Do a full desaturation on it.
According to this, I'll go to the pull down and add a layer node.
We'll keep the desaturation in the number two node, and then the number four node will
isolate the motorcycle.
I'll use the sampler tool to pull the qualifier color, do a little cleanup behind the seams,
and we now have got the shot with the red motorcycle on the grayscale background.
We can look at the original image and see that the color of the bike remains the same.
What's happened here is that the layer mixer has taken the red motorcycle, defined by its
qualifier, visible here, or here on the gear, or by switching our monitor to the highlight
black and white, that's option shift H, here on the menu.
As the alpha coming into the layer mixer, the layer mixer is taking whatever is inside
the defined alpha and placing it over or on top of the background image, which is the
desaturated image.
Now let's see what happens when I morph this to a parallel node.
Well that's unexpected.
Where did our colored motorcycle go?
All we have now is the desaturated image.
What happened to our red bike?
Well, when I first started out, this really confused me.
To my way of thinking, the two inputs of two images should be mixing in some fashion, and
that's the point I want to make.
Parallel is a variation of the serial approach, not of layer.
Let me put that another way.
As I said before, serial is a do this, then do that, then do that.
Parallel is a do this and that at the same time.
Now somewhere inside the parallel node, there's a secret sauce that determines the order and
priority of the operations.
When you look at it that way, you'll see that the first input is saying, desaturate
the entire image.
But the other input is simply saying where it's using the qualifier.
As you can see, there's no color correction, no saturation, nothing.
So this is basically telling the parallel node to do nothing.
So as far as the parallel node is concerned, all it's saying is a do a desaturation and
nothing else.
You've defined the area to do something, but you haven't told it to do anything in
that area.
There's a huge difference from the layer node.
If you actually want to achieve the red, you have to counterbalance or fight the competing
desaturation by upping the saturation in the qualified area, from 50 up to 100.
Then we get the combined image and again it matches the original.
But this is because it is simultaneously desaturating everything and super saturating
the area where the motorcycle is qualified.
And by doing that, the color has balanced out, has been pulled down to 0 by 1 node,
up to 100 by the other, so the balance is 50.
Parallel node has tried to serve two different controls, two different tasks, at the same
time.
But what if the commands are different controls?
Let's reset the super saturation back to normal and look at isolating this motorcycle
correction node using view highlight.
Instead of saturation, let's slightly shift the color of the bike towards blue by giving
the gamma and gain a blue nudge.
Before I go back to the result, just think about what I've told it to do.
I've told it to desaturate everything and, in the motorcycle area, shift everything a
little blue.
So what do you think it's going to do?
Yep, it made the motorcycle a little blue.
Not the red plus blue, the slightly magenta color we had, but it followed the instructions
to shift the bike area towards the blue, combining the instructions, not the imagery.
What would happen if I morph back to layer mixer mode?
I'll get the magenta bike on top of the background because remember that layer is putting the
image on top of the background.
While parallel, I was taking the instructions from this one and which is in turn a little
blue, an instruction from this one, which is to turn everything black and white and
combining them.
Now I actually have a lot of fun with parallel node.
It's like playing with watercolors to me and a great tool to explore the unexpected.
Let me show you something I did here just while I was thinking about it.
This is a lot of fun, taking the original image, took the overall highlights to blue,
used a qualifier to turn everything green into golden, and in the last one I ever so
slightly shifted the lowlights cooler.
The result is fun to me, getting a cool fall day.
I find that parallel is fun for just mixing stuff up, playing with possibilities.
It's a blast.
Play around a bit and I guarantee you'll come up with at least a few looks that you'll
want to add to your library.
Moving on, let's go back to our original setup here.
Well, does not care about the order of your inputs.
Let's swap the inputs around.
If I had asked for chocolate and vanilla, or vanilla and chocolate, I would expect to
get the same thing, and I'd expect it to taste the same.
And that is in fact what happens here.
Let's undo that.
Nothing changed.
Now let's try that in Layer Mixer, and go ahead and reverse those inputs.
Now here, all we get is the desaturated image.
What happened here?
Well, the full desaturated image went on top of the red motorcycle parts.
This is a really important point.
Parallel again is closer to Serial than it is to Layer.
While they both combine two image streams together, they in fact do very different things.
And occasionally, you can get them to do the same thing, or at least look the same, but
they're working in very different ways.
Serial and Parallel are two ways of applying color corrections to an image.
Layer is exactly what it says.
It is layering.
It is not combining color instructions, but it is taking images and stacking or compositing
one on top of another.
Let's finish by looking closer at the Layer node.
We'll take it back to our starting setup.
There are three important things to keep in mind when using Layer.
The first is the order of these inputs.
You can add inputs.
You can delete them.
As we found out, unlike Parallel, order is very important.
And you are making a layer cake.
You start here with your base layer or background.
Then you add your next layer, the red of the bike in this case.
I can add another layer by keeping one of my layers selected and choosing Add Layer node
from the menu.
In this case, it makes it add another layer to our stack.
Let's just add a circle window.
What you see here is the layering of the area defined by the circle mat on top of the other
two.
This one is higher, even though it is in the lower input, and is higher on the stack, overriding
everything behind it.
The second important thing is the alpha or mat in each layer.
In this case, the alpha looks like this, a circle.
For node 3, the alpha is the qualified or keyed motorcycle.
The background alpha, to be honest with you, doesn't matter.
It has no effect on the composite.
Let's get rid of that one.
So now we have order, alpha, and the final thing I want to talk about.
Composite mode, which you get by right clicking the layer node.
You see normal, add, subtract, etc.
We'll go over these in another tutorial, but each one of these creates a different combining
of the layers.
Just to repeat, the layer order will determine what goes on top of what, the alpha will tell
it where or what portion of your image to combine, and the mode will tell layer exactly
how to combine the images in ways that parallel can't even begin to approach.
One last point, layer allows you to combine images that come from the outside, such as
those you get by using the external matte node to apply scratches, for instance.
If you tried to do that with parallel, well, so far all I've been able to do is crash
the whole program trying that.
Thanks for hanging in there.
Watch it again if you think it'll help.
I promise the next one will be more fun and show the ways that you can use the layer modes
in very interesting effects.
Please subscribe to my blog, greymarshall.com, slash blog.
Follow me on Twitter, at greymarshall, or follow the Vimeo channel, vimeo.com, slash
greymarshall.
Or not.
Thanks for taking the time.
I'll see you next time.
