Back in February, I made
a string of videos on picture profiles, LUTs and grading
tips for the Sony cameras, and while the reception
of those videos was excellent, many of you did express some
concerns dealing with noise, especially with HLG. So today,
we're going to talk about reducing the noise in your
images and hopefully apply this to other profiles
and cameras as well. Let's get Undone!
[offbeat music] ♪ Gerald Undone ♪ ♪ He's crazy ♪ What's happening, everybody?
I'm Gerald Undone, and "I'm up all night
to get lucky." So, I recommended HLG on
those previous videos as a gamma that gives you a pretty good
balance between dynamic range, pleasing colour and a relatively
easy workflow, but based on your comments, it seems that many
of you are having issues dealing with noise when using that
profile, and after some further testing, I think I might
have figured out why that is. I tried testing
the same profiles that I tested before,
but this time underexposed, and sure enough,
HLG produced more noise than any of the Cinegammas or even
S-Log2 when underexposed. So, the solution there
is to never underexpose HLG, and in fact, overexposing
a little bit is probably best, and that goes
for most of these gammas. So, let's take a look
at how far you should overexpose these different gammas,
as well as how to reliably evaluate that
exposure while shooting. But, before we get into that,
if you've already botched the exposure on a shot
and there's no going back, then perhaps a clip
or two from today's sponsor, Storyblocks could come in handy.
Storyblocks has an impressive collection of stock footage
covering a wide range of subjects with unlimited
downloads and 4K footage. The interface is easy to
use and navigate, and the clips are royalty-free for both
personal and commercial use, so you can use them as much
as you want, wherever you want. So, if you think you could
take advantage of a fantastic library of quality
stock footage and effects, check out Storyblocks using the
link in the description below. Alright,
so I've set up a control shot so I can show you how
to dial in that overexposure, but the camera configuration
in order to be able to capture this is pretty funny.
So many clamps and arms everywhere,
I posted about this to Instagram in my story, which you would
have already seen if you were following me. But anyway,
there's three tools that are going to make this possible.
You can either use zebras, waveforms or false colour.
If you're shooting on a hybrid camera and don't have a
separate monitor, you probably only have access to zebras,
so we'll definitely focus on them, but I do highly
encourage you getting a monitor like the Ninja V, which also
has an external recorder, which is excellent when paired with
the Sonys not only because of the extra tools for exposure
but it'll only be easier to edit those files using those
Codecs like ProRes and DNx. Alright, so this, as it stands
right now, is a balanced exposure, meaning that
we've exposed to the middle. I'm using the standard HLG with
the BT.2020 colour, and we've put our middle grey at around
50% on the exposure scale. That's exposed to the middle.
This is not underexposed by any means and will usually
produce pretty great results if the image doesn't need
to be pushed very much in either direction in post. So, let's
look at how we can achieve this exposure using the three
tools that I mentioned. So first up,
let's look at zebras, which right now, are positioned
on the Sony screen, so you see how there's the striped
patterns on the middle grey? That means that our middle grey,
or at least that section of our middle grey piece on the
colour checker, is at about 50%. Now, let's jump over
here to the Ninja V, and we'll turn on
false colour. Now, false colour also
provides an exposure reading and it's based
on this legend on the side. I don't know how well
you can see that, but basically, the colours range from,
like, purple, through blue, then grey, then green
is right in the middle and then it
starts going up from there. A lighter grey colour
and then yellow-y orange, then red when you're
really overexposed. So, if we want something to
be at, like, 50%, then we want it to turn that green colour,
and we can already see on false colour that we're
getting a little bit of green down here on the middle grey
and the rest of the middle grey is just a little bit
less exposed than that. So if we want that
to match our zebras, then we can determine,
I guess, how the zebras are set for that
particular picture profile. So we can just come down
here and change our zebra reading by going down until,
I would say about [high-pitched] 48, 47.
[normally] 46, let's say 47. 47 has a pretty similar
result to false colour, so that's going to be
exposed in the middle, using 47 zebras for HLG, or we
can use false colour over here. And the other option
that we have is with waveform that I mentioned, that's the
third tool. So here we can see that our middle grey bar is
reaching right about 50%. So those are your three
different ways of confirming exposure, or in
this case, confirming that you're exposed to the middle.
Now, I'm just going to record a small sample of this one
so that we can come back to it later on, and we'll refer to
this as middle exposure when we evaluate these
in post for comparison. Now, let's underexpose the image
by bringing it down two stops. So, I'm going to bring
it all the way down to ISO, so that's one stop,
and we can't quite go down two stops because the
lowest it goes is ISO 125. We can stop it down just a notch
with aperture to make that. So, we have two stops down now,
so we're at f/2.2 at ISO 125. And now we can see that our
middle grey is way down here, probably at like thir-- 28,
30%, somewhere around there, and that is grossly
underexposed for HLG. And if we take a look
at what that looks like with the false colour on,
well, you can see that most of the image now is blue and dark,
and that means we're not, we're not doing well here.
So let's do a quick recording of that, so again we can
compare it in post later on. Now, at first glance,
you might think that this would produce less noise
because we've lowered our ISO, and I think that could
be potentially where some of the people that commented were going
wrong, because even though, yes, our ISO is all the
way down to the base, when you correct for this
in post which will require quite a bit of boosting the signal,
you're going to bring all that noise back, plus you're going
to introduce other problems like colour shifting
and some weird artifacts. Now, let's shoot an overexposed
version as well to compare, but the trick here is
how far to overexpose? Now, there's a few factors
that you need to consider when doing this.
So first, you have to consider the capabilities of
the profile that you're using because some will have a lower
maximum output level than 100%, for instance,
with HLG 1 and 2, they're lower than
HLG and HLG 3. Both HLG and HLG 3, the maximum
output level is set to 100%, but HLG 1 is set all the way
down to 87% and HLG 2 at 95%. And these will be
reflected in the waveform if I bring that up here.
Your widest points, so if we increased the ISO
all the way up, do you see how, once it gets to 100, then the
line becomes flat across? That means it's like
clipped beyond that. There's no longer
any detail in there. Well, that straight flat
line would occur lower with HLG 1
and HLG 2. That's how the maximum
output level is affected. But the curve also matters too,
because S-Log2 for instance, which has a 106% maximum output,
actually can retain a lot more in the highlights and has more
overexposure potential compared to Cine 1, which has a maximum
higher output level of 109%. Because it's more important
to curve there, the of S-Log vs. Cine 1 than
it is just their maximum output level. Now, you can see
this in action just by counting your stops while watching
your waveforms or your zebras, but it's important to know the
clipping point of each profile, otherwise, you won't
know how to set your zebras. Now luckily, most of the
manufacturers include their clipping point in their manuals,
and often, if you don't really know a safe one,
it's usually, probably, a 100%, unless you're using
a specific profile like the shorter ones like I was
talking about HLG 1 and HLG 2, or like V-Log
for Panasonic or S-Log3. They all have lower than 100%,
but if you're using more standard gamma,
a 100% is probably the safe one, but even
if it is a 100% like HLG, you should probably still
set your zebras lower than that. So, we're going to set
our lower limit here to 95+. The reason why we want to give
ourself that extra bit of buffer is because just because the
clipping point is 100 doesn't mean that all of the individual
colour channels are going to come in at the same potency,
and so you may have a white image that isn't clipped,
but the red channel might be clipped a little bit,
or the blue channel. So, it's safer to give
yourself that extra 5-7% buffer so that even if
the white's coming in at 95, maybe the red is coming in
at 99, but it's still all safe. It's still all
under that 100%. So now that we've set
our zebras, which we've set them to 95 as we can see here,
and we'll leave the waveforms up on the side of the screen as we
increase our exposure using ISO, eventually the zebras start
to creep in, and we can see them there on the white, and now
if we look over here on the waveforms, we can see that we're
just almost touching the 100%. We're not up to 100% all the way
because we set them to 95, so we have a little bit
of that safety there, and if we compare this
instead to our false colour, take the waveforms off,
we can see that the yellow is now starting to turn
into that orange and red. Now, let's overexpose it
clearly, and you can see how this white bar here turned
completely red, and if we look over here to the zebras on this
screen, we definitely have them in several places on our image,
so this would be overexposed. So, we're
going to bring this down. You might be able to get
away with this, 'cause there's only that little touch there,
and we do have it at 95, so we're probably safe,
but if really wanted to be safe, you could bring it
down one notch further, and that has to do with
preserving details. So, even though you are
still maybe in that safe range, sometimes it's better
to come down one more notch because as
you squeeze the highlights, usually with these curves,
there's more, more information sort
of crammed into the top end. So as you, sort of,
cram all that information into the top part
of the curve on the highlights, you're going to leave
less space to actually pull that detail out when
you bring the image down. So, it's usually smarter
to save yourself a little bit space for the clipping
and then save yourself maybe one more notch to
preserve highlight detail. If you have no interest
in the highlights, like in this shot,
there's a black background, but let's say that it
was some already impossible sky, well then I guess it
doesn't really matter. But, if it was an image
where you were trying to expose a little bit with those clouds,
well, then there will be a difference between that
one notch or the extra notch because you might lose
a little bit of the detail on the fluff and the
cloud if you went up, so bring it down one
more notch from there. So, for HLG, a safe point
here for overexposure from the middle.
We'll talk more about overexposing based on the
manuals, but we're talking about from the middle. Yeah, we have a
good middle exposure here at 400 or even, even 320.
320 to 400. And now,
we can safely overexpose up to 640 and probably,
800 would be OK for most purposes, but if you
wanted to be safe and still have a little bit overexposure,
we'll say 640 is the safe one. So that's 2/3rd to one full
stop overexposing to the middle. And this whole process is what's
known as exposing to the right, but there's different
interpretations of that. Some people believe it means,
you know, jamming your histogram all the way to the right.
Other people believe that it's moving it to the right
but still allowing for roll-off. This is the way that
I expose to the right. It's not all the way to the
right, but it's as far as to the right as I feel comfortable
with when it comes to trying to preserve most of the image,
but still, you know, working in a decent
amount of noise reduction. Now, the other thing
you need to consider when you do this
is your subject. In this case, our subject
is this colour chart. So, we're obviously
overexposing correctly. But, like I mentioned before,
in that other scenario, where maybe your background was a sky,
and maybe the sky is important, you have to ask yourself
what the subject is. If it's, you want to have our
colour chart, let's call that, sort of, our secondary subject,
if that's supposed to be, kind of, in the foreground,
and you're supposed to know that it's there,
what's really important is the sense of what the
sky means to that scene or, you know, they're looking
up at the sky or something, then you, maybe, don't want
to use this based on the chart. You'd want to use based
on the sky so you'd want to do this, you know,
expose the sky until the zebras then come
back down and then maybe come down one more if
detail is important to you. And then maybe when you
looked down at the chart, you found that it's quite dark.
That's OK because you're going to leave that chart dark.
What you don't want to do and what I see happen a lot is
that people will do that thing where they expose for the sky,
they retain the sky, but then they also bump the
shadows to make the chart also be naturally exposed,
and that's not really possible. You'd have to light the chart to
increase the exposure on it. I'm talking, like,
like a unified light experience, like just the sky or a
natural light environment. You wouldn't be able to expose
both of those things, so you shouldn't bump the shadows
because now your chart is going to be really, really noisy,
'cause you exposed for the sky. You would have to light
the chart and increase the exposure on it until
it followed the same ETTR principles that we're
discussing in both scenes. If the two elements
have different exposures, you can't ETTR to both of them
successfully at the same time. So then your other option
that you'll see a lot in films is that the subject will just be
exposed for, and then they just let the skies blow out, because
who cares about the skies? We're looking at the
expression on the person's face in that war scene or whatever.
But anyway, there's two different things to keep in
mind, make sure that you're applying these principles to
what is important to you in the frame. Not two things,
and not the wrong thing. I'm going to give you a
little bit of a cheat guide here because it's quite possible that
you're not going to be able to do all this every time you want
to set up a run and gun shot. If you have the ability to set
up a complete exposure situation like this with a monitor
and the correct tools and the colour chart and everything, I
would recommend that you do it. But if you're running
around and you're just trying to shoot some subjects of skin,
I'm going to give you a couple of percentages
to target here so that maybe it can make your life
easier just with zebras. So, if we look at the
colour chart over here, we've got
our different skin chips, and they range from very
fair skin to darker skin. So, based on whatever
subject you're shooting, I'm going to give you
some percentages so that you can get a rough idea of where
to put their skin since they're likely your subject in order
to get the correct overexposure. So, to do that, we're
going to go into our zebras, and we're going to change
our standard range, and we can dial this into
whatever skin tone that we want, to whatever
exposure that we want. But first, let's see what the
different skin tones are by just increasing our zebras until
we see where things are placed. So, let's go all the way up here
to, till we get to the top chip. So there,
78. You see, usually, people try
to remove the zebras because it is an indicator of overexposure,
and you want to remove anything that is overexposed, but you can
also use it when it's your only tool, as in the case of
the Sony mirrorless cameras. You can use it as a 'tell me
the exposure of that subject', so right now, our subject,
the top chip, is 78%. So, if we set this at 78%,
and then we ran around and filmed our subjects,
and they had similar skin tone, when we saw zebras on their
face, we would know that we would have them at 78%, and that
might be what we're looking for. So, I'm going to go
through and show you the different
percentages for each one. So that's 78%
for the top chip. Now, we'll move down it's
about 76 for the one below it, then we'll move down and that
one there... [audio fading out] So, hopefully, that'll give you
a little bit of a cheat guide to know roughly where to set the
exposure for those different skin types, and then once you
have that, you can switch to your second zebras down here,
which was set to 95% just to make sure that
nothing is clipped. Now, these percentages
only apply to HLG or HLG 3, they're both pretty much
exactly the same. Like I said, I wouldn't really monkey
around with a HLG 1 and 2 unless you want to completely
change that whole workflow and with HLG and HLG 3, it
doesn't really matter, they're interchangeable. Just one is
more saturated than the other. But, before we move on to the
other profiles where I'll give you the exact same, sort of,
cheat guide, I want to talk a little bit about other ways to
affect the noise in your image. Firstly, as I've recommended
with all the profiles, is to set your detail
in the camera down to -7. This will remove any sharpening
that's applied to your image because usually,
when you apply sharpening you'll also roughen up
the edges of your noise, and it makes it more noticeable,
but some of you seem to still want to apply some sharpening,
and that's OK, I guess. I don't recommend it.
I think you should always do sharpening in post, but if
you need to sharpen your image here in camera, there's another
option under Adjust called Crispning and what this does
is affect, it's sort of, like a tolerance level. The higher
you set it, the more contrast is required in order for the
sharpening to even be applied. It affects
the way that that luma is, like the luma changes
between two areas. If we were to change this
mode to manual so we can adjust it and then go down here
and increase the Crispning, so at zero it's willing
to apply the detail changes, the sharpening to a much wider
range of elements in your frame, but if we increase
the Crispning all the way up to seven, well now it's only
going to apply that sharpening, the detail settings, to
the most severe changes in luma. So this shouldn't affect
noise that much because noise is generally not very contrast-y,
it's pretty like pretty even when it comes to luma, so if you
have your Crispning high, then you can probably
still have some sharpening in your image without worrying
about it affecting your noise. But if you have your detail set
to -7, and this doesn't really matter, because it's not really
sharpening being applied anyway, but just something
that you should know. Alright, now let's blast through
the other gammas and make some determinations about
their overexposure potential, because the greater the
overexposure potential of gamma, the greater the noise
suppression potential of that gamma as well,
because in post when you bring your image back down
you push the noise down with it. So, the winner here is
definitely going to be S-Log2. Now, S-Log3 is another
really great example, but I don't really recommend
using that on these mirrorless cameras because
the codecs in them is not really built for S-Log3,
that's more for cinema cameras. So, with what we've
got to work with here, S-Log2 is probably the best
bet for overexposure potential, and it works pretty
much the same way. So first, let's go ahead and achieve
that middle exposure. Now, like I pointed out earlier,
I've already found that around 45 on the zebras
is about the same as false colour and waveforms, and if we
put our waveforms on, we can see that yeah, it's pretty
much around the middle there for middle grey.
Now, it gets more complicated than that when it comes
to S-Log, and that's because remember how I said
you should check your manuals, so you know what your
maximum output level is for your different gammas?
Well, if you do that for Sony, it will tell you that when it
comes to exposing S-Log2, if you want the maximum dynamic
range, that you should set middle grey to 32%. This will
give you the most dynamic range, but it doesn't exactly
give you the best image, and other people have
covered this in-depth already. Wolfcrow has a whole guide
on it, Alister Chapman's talked about it many times,
but it's important to know this because when those guides
and other people talk about overexposing S-Log,
they're usually talking about overexposing from
that 32% point, not from exposing
to the middle. So if somebody said, "Oh, I've
overexposed two stops on S-Log", two stops after achieving that
32% is usually what they mean. So let's start by putting
S-Log down to 32% middle grey and see what that looks like
for underexposed image, and then
we'll move up from there. Now, we can't really lower
our ISO anymore because we're at the base ISO for S-Log,
which is 800, but we can close down our Iris until we get
down to about 32%, but in order to confirm that we are at 32%,
as we've talked about, we can use our zebras,
So we'll set this down to 32, so it might be harder for you to
see the zebras and the grey when our exposure is this low,
but they're starting to show up around 3.2, and they're
also there around 3.5. And based on our waveform,
I would say that f/3.5 is probably
around 32%. So this is our correct
exposure according to Sony, and if we look at our
finished image here, we can see that it's really
quite dark and remember what I was telling you about
all the blues and blacks? We can see those all over
our false colour, and so this is going to produce an image that,
when you correct it up to the levels that you
want it to be in post, yeah, maybe you'll get
better dynamic range, but you're going to get
a whole lot of noise. And now, let's move it
up to exposed in the middle, and then we'll also
overexpose it, and we'll keep track of how many stops
there are in between there. So right now, we're at f/3.5,
ISO 800, and now we'll start increasing the exposure,
and right away at f/2 ISO 800, I would say
that we've achieved it. We have our zebras right there,
and we have green over here, and if we look at
our waveforms, yeah, we're right about
in the middle there. So we have our middle exposure,
and that required us to go from 3.5 to f/2, so that's
1 and 2/3rd stops, so just to, sort of, get to the same middle
exposure which has pretty decent noise performance, we have to
overexpose by 1, and 2/3rd stops from the manual recommended
S-Log exposure level. And now, let's overexpose it
by applying the same principles that we talked about before.
So, first of all, like I said, remember you have to, you have
to know your clipping point and then give yourself a little
bit of safety room there? So the clipping point
for S-Log2 is 106%, but I found that setting
it to 103+ is good for that first,
sort of, a bit of safety, and then I usually go
one stop down from there. So, we'll set it to 103,
and we'll go all the way up until
we start to see zebras. And there, ISO 5,000 is when
they first start to come in. So we'll go down one more notch
from there, and that's ISO 4000. And, if we check this
over here on the waveforms, we can see that
we're above 100 because again, S-Log2 can go up to 106,
but remember that flattening of the line
that I was talking about? We're not getting any
of that because our image isn't really clipped.
If we go up higher, though, do you see
how now the line just gets flat? So that means it's clipped.
So this is the maximum that I would ETTR, or expose to the
right of my S-Log because now we're two full stops above when
we were exposed to the middle, we were at ISO 800,
now we're at ISO 3200. And that puts us over
three stops overexposed from the 32% manual
recommended exposure. And in that Alister Chapman
article that I was talking about, he tested one, two
and three stops over the 32%, and found that by the time you
get to about three stops over, you do start to lose
some of that highlight detail. It does start to become,
you know, flattened out a little bit, so he usually recommends
aiming more for that two stops, you know, area overexposed,
which for us, that would put us probably at around,
anywhere between, I would say, ISO 1600 and ISO 2000
in this example here. So that would look something
like this to this, but we'll do a recording at both,
and we'll compare to take a look at what the zebra
levels are for that. So, the equivalent level of the
HLG overexposure, which we said was around 2/3rd - 1 full stop,
on this one over here is double that it's, like,
1 and 2/3rd stops from middle, and like 2 and a half
stops from, like, 32%, but lastly let's just
blast through the cine gammas so we can see if there's
anything to glean there if one is better than the other
for overexposure potential. So, let's
switch over to Cine now. This is Cine1, and all these are
going to be using the Cine colour, but it doesn't really
make a difference. So now we've got our middle
exposure with Cine1, which puts us at f/2 and ISO 250, and if
we check our waveform over here, yeah, we're coming in the
right in the middle, so perfect. So now, we'll just see
how many stops we can overexpose with Cine1
before it starts to clip. Now, Cine1 has a maximum
output level of 109%. So, let's go ahead and give it
its full capabilities. Like I said, you probably don't want
to do this. You probably want to give yourself, like, at least
a 5% leeway there for channel clipping, but let's go ahead
and do that and just see how much we can go. So right,
ISO 250 so go one, two, three, four, so five,
so one and 2/3rd of a stop and we're clipped where,
where whites are blown out. We go down one, and we're coming
in pretty hot over here on the false colour, and you can see
the zebra starting to creep in, so I would consider this one,
you know, the danger zone, so I'd probably go down one more
from there, and that's ISO 500. So we really only gained
one stop that's our maximum, our maximum overexposure potential
on Cine1 is only one stop. So despite it having 109%
of its maximum output level, it's really only achieving
the same, sort of, overexposure potential as HLG,
but without that, without as much of the
highlight squeezing, so you'll see that your
white chips on Cine1 seem really bright compared
to what they were on HLG. Uh, Cine2 is pretty much
exactly the same as Cine1. It just has a lower point.
Instead of 109 maximum output, it's 100% maximum output.
So we'll set that to Cine2, and everything else should be
pretty much the same. We just need
to bring our zebras down. One thing that you'll notice is
that if you ever set your zebras in a Sony camera higher
than the maximum output level of the gamma allows,
they just won't work. So, for instance, it's set to
107+, Cine2 can't go that high. It doesn't matter how
high I overexpose, the zebras will never appear. So we have to
set them within the maximum clipping point, which is why
it's important to know what the maximum output level
of that given gamma is. So let's go ahead
and set them to, well, set it to 95,
just like we did with HLG. 100 is where they go,
but 95 will give us that 5% buffer zone there, and so let's
bring this down. [chuckling] So that's 1/3rds, 2/3rds,
and we're already introducing some
clipping at 95, and then by one full stop over,
we're blown out. So if you want more overexposure
potential, Cine1 has, probably, a full half stop
better than Cine2 does. So now we'll switch
to Cine3. So we will increase our zebras
up to the maximum output level, see they disappear at 109,
so that's obviously out of the Cine3 realm,
so we'll go down to 108, and the zebras are there,
so we know that we're working and we'll just bring these down,
and so they start to come in around ISO 1000, so 800 would
be the maximum that we could go. Now that is a bit higher
than what we get from Cine1, but let's make sure that
our middle grey is the same. So let's go up to here
and set this to 45, and then we'll jump this
down until it starts to appear and yeah, it looks the same,
but our image is darker. At ISO 250,
it's darker than it was at Cine1 and Cine2,
and that's part of Cine 3. It's supposed to, you know,
work with the shadows a little bit more, so right now,
if we increase exposure from here based on
the same lower limit, we can go 1/3rd, 2/3rd,
a full stop, 1 and 1/3rd, 1 and 2/3rds, and then
two full stops when we come in. So Cine3 is actually
giving us more latitude than Cine1 and Cine2 was,
so based on this purpose, if this is your intention,
Cine3 will probably give you a better image overall, but there
is one thing to consider. Cine3 is darker,
so again don't fall into that trap where you try to boost
the shadows because that's just gonna add the noise back,
and you're gonna end up with the same result as if you'd shot
in Cine1 in the first place. Cine3 will give you quicker,
better noise if you shoot this way, though if that's something
that's important to you. Now lastly, we'll look at Cine4,
which again is the same idea but even more extreme
than Cine3's version. So let's go ahead
and change this to Cine4 and first off,
let's find our middle grey. So we'll come down here,
and it's the same again. 250 is the same,
so our middle grey is the same for all of them,
it's just, you know, how dark our shadows are getting
and how much headroom we have. So now we'll go, first of all,
let's overexpose so that we can see what level
our zebras will work at. So we'll go here,
108 works, 109 works. So, interesting how Cine3
doesn't have the 109 potential, but Cine1 and Cine4 do.
So now, we'll set it to 109, and let's put this down to,
we'll go back to 250 and then we'll
walk ourselves up from there. So 1/3rd, 2/3rds, full stop,
one and 1/3rds, one and 2/3rds, two stops, two and a third,
two and a half is no good. So two and a third.
So this obviously suggests that Cine4 has the best
potential of the cine gammas for that noise reduction
aspect if that's what's important to you because you
have the most overexposure potential even though you
do get a pretty extreme contrast with Cine4, so that's
something to keep in mind. So now, let's load all
these up in Resolve and examine the differences. Actually,
before we do that, let's take a quick look at the movie gamma,
just for fun, in case any of you shoot that way.
And by the way, the movie gamma is the same gamma as if you set
the picture profile to off. When you're in photo mode,
off is the same as a still gamma,
and when you're in movie mode, off is the
same as the movie gamma. So let's set it to that,
and we'll go through the exact same procedure here.
So we'll find our middle grey, which, perfect, it's still the
45% zebras on Sony still seems to be working for this purpose,
and then we'll just see how much overexposure latitude we have.
Look at the contrast on this thing!
So, we'll set it to... We'll do 95+ just to be
safe and-- [chuckling] I haven't even changed the
exposure, and 95+ is already showing as overexposed, so
there's nowhere to go with this. So all that that tells you
is that if you want to use movie gamma, you better make
sure that you nail the exposure. It's the easiest one
it's pretty much completely finished
as soon as you record it. So there's no post-work
involved, but exposure is crucial, and that's going
to be a common thread here. The ones that require the
most work in post give you the most latitude to be able to
monkey with your noise or pull your highlights back in,
and the ones that require the least amount of work in
post generally require the best efforts when it comes to
exposure, because like I said, this is funny,
I have to actually underexpose a notch just to stop the
highlights from clipping, which now puts my middle
grey below the middle point. Alright,
so I've got the clips loaded up here in Resolve and now
that we're done shooting, I can turn off that lantern,
so the right side of my face isn't overexposed to
the point of losing detail. Anyway,
so in here we've got them all set up the first
three clips here are HLG. We go under, middle,
over and then, we've got S-Log. We've got under, middle,
slightly over, a lot over. And then we have that Cine4,
you can see the Cine4 right away
it's so contrast-y. So let's start with
the middle exposure HLG, and we'll set this one
up to be our control, and then we'll make
all of them match that. And this is easy to do
because we're using a ColorChecker,
so we can just position our bars where they
belong on our waveform. So it's going to be the
exact same principles we did when shooting, but now
we're just going to finesse it a little bit in post, So if
we go ahead and we look here at the waveform in the bottom
corner, we can see that our middle grey is exactly where
we want it to be, right on that middle line, which is
represented by this 512 here, and our white is pretty good.
It's, it's a little bit down, and that's what you get
with Log. It kind of pushes the whites down a little bit.
So the first thing we're going to do is we're just going to
increase the gain until our white bar reaches around that
90% point, which is about here. This is at 90% whites,
that's where we want it. And then,
we'll see that our middle is a little bit higher now
because we increase the gain, so we'll just lower the
gamma off, and we'll just sort of finesse these up and down,
gain up and gamma down, until we get them to where
they each belong where they do. This one at around 90%,
and this one around halfway, and then we'll just look
at the image and see what the contrast is like,
and see if we're happy with it at this point, we can
make a change with the lift, if we wanted to, you know,
bring the shadows up or down. Now, like I said,
bringing the shadows up is going to be problematic for noise,
so usually, if you exposed well, it would
only be down or nothing. And where we are right now, I
would say it looks pretty good. This absolute black area here is
coming in right around where we want it, which is in that,
sort of, like, 6-10% zone, and so I would say this image
looks fine, and this will be, sort of, our baseline that we'll
get the other ones to match, So now let's jump over
to the underexposed image, and we're gonna have to
boost this one up to match that, but we have
the same guide to do it. We're just going to hit those
same points, so let's start by increasing our gain
until we get to that 90% point. And now,
here you can start to see that channel separation
thing that I was talking about. If we switch to the Parade,
you can see that the red is actually higher than the green,
which is higher than the blue, so when we go back to our
waveform, we've got different colour separations in those
bars, which is why it's possible to clip one channel and
not the others which is why you want to leave yourself
a little bit of a buffer there. Here, we find that our gamma
is still too low, even though, like our mid-tones aren't
hitting that halfway point, even though our gain
is all the way up, and this is where we're gonna start to
get into noise town and when we're also going to shift our
colours in an unwanted way. So let's increase our
gamma until we get that to about the halfway point,
and then we can, you know, bring our gain maybe
back down a little bit, gamma up a little bit,
just, again, balance them out. And then now,
you can see that we have a lot more noise down here because
our black area is sitting up a little bit higher, but look at
all this junk around here. So we're going to lower
the lift a little bit to bury some of that noise
and match the overall look. So we can come down
a little bit more. Do you see how it's
just less impactful now? The, the colours just
seem more washed out. That's another thing that
happens when you have to lift everything up, and then
if we look at that noise, this noise here is what I was
talking about when I said if you underexpose your HLG,
you get a much dirtier image. So let's
play this one for a second. You can see that the noise is
quite aggressive, and if we switch to the one that's
exposed in the middle, you can hardly
see it at all. Now, let's go to our
overexposed image and bring that down and see if
we notice even more improvement. Now you can see here that
we're already at that 90% point for our white, so we're
not going to have to increase our gain, which is
going to be great for removing noise because we don't have to
increase anything, we can just lower things. And our
gamma is going to come down because our mid-tones are quite
high, so let's bring our gamma down until it gets to about
halfway, and then we'll just, we'll increase our gain a little
bit just to make sure that we don't lose that, sort of,
like, the intensity of the white but other than that, I would say
we're pretty good already. And you can see now how
thin this bottom line area is, that's where our noise
is going to be hiding. So, if you look at the thickness
of that and compare it to the thickness of this one,
it's a little bit thicker and then compared to thick
this one, this is when it takes up like the
bottom 10% of the image. So now, let's just compare
the noise between all three. I would say that you're
definitely making a big step going to the middle exposure
over underexposed, and then you just get a small advantage
going to the overexposed. So let's compare the
underexposed S-Log to the underexposed HLG.
Where on the underexposed S-Log, yeah, it's definitely noisy,
and it's thicker than the overexposed one,
but it's not near as bad as HLG. So now let's go up to
the ISO 2000 version, which was like two
and change stop over the recommended 32%,
and it is a little bit cleaner. But, just like we expand
with HLG, the difference between middle exposure
and slightly overexposed isn't as huge as underexposed
and middle exposure. Alright, now, lastly,
let's take a look at the Cine1, just to see what's going on
here. Don't be alarmed, I know it looks like it's clipped over
here, but that's because we were exposed above 100%, but we can
bring that down with our gain and then, as you can see,
everything comes back to normal. So we'll set that down to
around the 90% zone, and take our gamma down just like we do
with all the other ones, the exact same technique, and that
is very, very, very clean. Let's compare that to
our overexposed S-Log... Similar, yeah, I mean,
the Cine4 is probably a little bit cleaner.
I mean, it's got to be because of what it's doing
to the bottoms there. So let's wrap up.
In my previous videos, I suggested that HLG is a great
choice for a well-balanced profile, and it certainly is,
but anything that's well-balanced usually isn't the
strongest in any one area and in the case of HLG,
it performs horribly when underexposed. So if you
plan on taking advantage of HLG, make sure you aim
for that 2/3rds of a stop to one-stop overexposure
to keep your images clean. If, however,
you plan on finessing your image is a little bit in post
and don't mind the extra work, then S-Log2 definitely has the
best potential for the lowest noise because it has the highest
tolerance for overexposure. Just make sure that
you're not overdoing it because, as we saw here,
there's definitely diminishing returns when it comes to noise
reduction via overexposure, and you're probably better off
bringing it back down a notch or two in order to preserve
those finer highlight details. So I've shut up, set,
so I've set up, sho-- [hissing] So I've set up a
control shot so I can... But that's gonna be it for me.
I hope you found this video entertaining or at least
helpful. And if you did, make sure you leave
it the old thumbs up and consider subscribing
if you haven't already. But if you did not find this
video helpful or entertaining, then what
are you still doing here? I would have ducked out
a long time ago. Alright... I'm done.