Today's video was primarily
intended to provide proof to a white balance misconception
that I encountered on Reddit, but I figured,
hey, while I'm at it, I might as well make a video
about it for you guys as well. And I'm also gonna use this
as an opportunity to address a somewhat-frequently
asked question I get regarding why I wear this grey
shoelace around my wrist. Let's get Undone. [offbeat music]
♪ Gerald Undone ♪ ♪ He's crazy ♪ [cryptic techno music] ♪ [slo-mo] What is-- [normally] --happening,
everybody? I'm Gerald Undone, and today we're talking
about white balance. More specifically, when it's
important and when it's not. So here's the debate:
I stumbled upon a thread that suggested that
it's not that important to set your
white balance in-camera if you're planning on
doing post-production, because you can always
change it afterwards. This obviously continued
with comments suggesting that that was only applicable
when shooting raw, while others insisted it
was possible with any format, and then some people said no, it's only JPEGs that don’t work
and video is fine, and some people said
video is only fine if you shoot 10-bit--
and it went on from there. Now, I realize
for many of you, the answer for this video might be
something that you know already, but even still,
I'd like to provide you with some tests
that you can use for reference. OK, So let's start off
with raw. Now, I've got
some raw photos here, but what I'm about to say
is also true for video if you have a camera
that shoots raw video. However, when I say
“raw video,” I mean proper raw. There's many raw light codecs,
like ProRes raw, that won't offer these results,
and thus your white balance can't be changed as easily
in post when using them. But again, this will depend
on the raw light codec you're using, and as always,
your mileage may vary. All right, so here
I have three raw images taken with drastically
different white balances. They all have a grey card
in the shot that we can use for setting
the white balance in post. And some of them
are overexposed and some of them
are underexposed. I did this intentionally,
and we'll talk a bit more about that
a little bit later on. But first, let's establish
what this image is supposed to look like. So over here I have two images. The first one was shot
using a custom white balance in the camera using this
grey card, and the second one was shot
using auto white balance. Now, you can see that
the auto white balance, you know, did a decent job, but it didn't perfectly get rid
of the green tint. And the light that this was
shot under sort of trips up auto white balance
systems all the time because it has a little bit
of green into it. So what we're gonna do first
is we're gonna use the dropper, and we're gonna click
on the grey card just up in this top corner
here just to set it and see how that matches the first
image, and as we can see, they're pretty
much identical now. So now we have a baseline
moving forward, and we also know where to
dropper the grey card to test and see if the white
balance will be conformed. Now let's go ahead
and apply that to the images with the manually set
white balances. So the first one here is set
to 5500K, so it's relatively close to what the lights
were running at, but again, it's gonna be
a little bit off. So let's dropper it up here... and I think we can say that
that's probably pretty darn close to what the custom
white balance setting was, but we can just confirm that
just to make sure. So... yeah, it looks
pretty much identical. OK, great. So we're gonna use
that going forward. Now, this one--
let's move forward one. This one was shot
at 10,000K, so it's crazy warm. And we'll use the dropper
and we'll apply it up here. And again, it looks
pretty much identical. There's the previous one,
and there's this one again, so I would say
that’s a great result. And then we'll go forward:
this one was shot at 2500K
so it's crazy cool, and we'll dropper
in the same spot, and again... you know, you can't really
tell the difference at all. If we zoom in, even,
and look at the colours here, I would suggest that-- yeah,
like, they're pretty much identical
right across the board. So this test confirms
that yes, raw images and photos are able to have their
white balance changed drastically in post
with no discernible losses. Now let's reset
these three images, and I'll tell you
why I wear this grey shoelace. When it comes
to making these videos, shooting the B-roll for them,
or even my product photography, my hands are often in
the scene, and I don't always remember to put out
a ColorChecker or a grey card. But if I have something
consistent from scene to scene, like this shoelace--
which happens to be a very neutral grey--
that I can anchor to to set the white balance off of,
I find that I get a much more consistent
look throughout my shots. I'll demonstrate this using
those same three raw images. So if we go over here--
this is the 5500K one again, and if you grab the dropper--
and I'll just set it off of this highlighted
part of my shoelace here. There's
the white balance there. Let's jump over
to the 10,000K one. Try to find the same spot. And one more time
with the 2500K one, and again,
try and find the same spot. And as you can see,
not only is it consistent between the shots,
but it's actually pretty close to the previous set
where we used the grey card. Sometimes I think it might
even work a little bit better than the grey card,
and I think this might be because of how close it is
to my skin, so when the light bounces off my skin
it casts a little bit of that colour onto the bracelet,
which then gets white balanced. And I think sometimes it creates
a more accurate skin tone. Now, I'm not
suggesting that everybody should go out
and get a grey shoelace and tie it around their wrist
in order to take photos, but having something
with you at all times that is a neutral grey
can be quite beneficial to keep a consistent look
throughout your shots. Just make sure that it doesn't
have its own colour, though, or that defeats
the purpose. Just so happens
that this grey shoelace is working
pretty perfectly for me. Now let's move on to the JPEGs
and try this test again. Alright, so if we start
off with the 5500K JPEG and do the exact same thing,
white balance up in the corner here, we can
see it stays pretty close. The shift-- you know,
from maybe 100 or 200K isn't a big deal.
It added a little bit of warmth and just one point of magenta,
and everything looks fine. But if we try this same
thing on the 10,000K image, the results are much worse. At first glance it might look
like it did a pretty good job, but if we look
a little bit closer at the colours here,
we can see that compared to the other one, we're losing
a lot of intensity there. And that's because when we
pull the warmth out of the JPEG in order to balance the white,
we're also pulling all of the intensity
out of our warmer colours. And this isn't something
that can just be fixed easily by upping the saturation,
because even if you increase the saturation in order
to get the oranges right, then when you compare to
the other image you can see that it looks completely
different because all of the colours
are now too saturated. So instead you'd have
to go through individually and tweak
the colours using HSL. Now let's try it on
the 2500K image... and, yowza!
That is terrible. So there's two major
problems going on here: the first one
is that it's overexposed now, and the second problem
is that we obviously have a terrible purple issue,
which is the inverse of what just happened
with the warmer image. So in this case,
because we're adding warmth, we're now contaminating
our blues and making them more purple. But this is why I have
two images, like I said, that are overexposed
and underexposed. So if we try it on the image
that's underexposed and do the exact same thing,
we can see that it looks a lot better as far
as exposure's concerned. It's still crazy out of whack with all the blues
turning purple now, and the whole image
has like a... you know, kind of purple cast to it,
but the exposure's better. And this makes sense
if you think about it. If you picture a white light
as really just being intense versions
of all the colours combined, then removing some
of that colour will make the white less intense,
or darken your image. And the reverse is true,
as we see here, when you warm up an image
by adding more warmth and more magenta,
the image gets brighter. In the case of this one
that's underexposed, it got-- you know, I would say
the exposure's a lot better, but at the cost
of corrupting your blacks with a lot of purple. So this test
has two conclusions: one, extreme white balance
adjustments are not good for JPEGs. Small adjustments,
like a couple hundred kelvin, is gonna be fine, but going
from one end of the spectrum to the other is gonna
have a terrible result. And two, it doesn't matter if
you overexpose or underexpose, because either will
cause problems depending on whether you're warming
or cooling the image. But if you do know
for sure that you're gonna be warming the images up,
then I guess you're better off
shooting a little bit darker. But obviously
the best thing to do is to set the white
balance while shooting. And the best way to do that is
to set a custom white balance in your camera
before you start shooting now, I'm not gonna give you
a tutorial on that because it depends on your camera,
and there's lots of different ways to do it, but usually it
requires you either take a photo and then sort of scan
that photo in the camera, or to set it live,
they'll put a little target, you hold up a grey card, put
the grey card on the target, press “OK,” and then
it sets the white balance. This is usually found
in “White Balance” under “Custom White Balance.” The second best option if you
don't have a grey card would be to use some other neutral
object, like my bracelet or, you know, a piece of paper
with 18% grey printed on it. And it's also handy
if you take a couple test shots or some frames of video
with that object in the scene so that you can use it to check
white balance in post later on. The third option, if neither
of those is available, is to eyeball it
and/or set it off your lights. So if all your lights
are set to 5600K, then set your
white balance to 5600K. But I would still probably
fine-tune it after that because sometimes cameras and
lights have different opinions on what a particular kelvin is
when it comes to white balance. And this is a lot easier
to do than you think. You basically just have
to get behind your camera, look at your screen,
be looking at your scene, bring up your
white balance setting where you can dial
in the kelvin, and then just look at the
scene and look at your screen. If you find that,
“eh, it's a little bit warmer than I want,”
cool it down a little bit. It's a little bit too cool?
Warm it up. Try and get it as close as
you can just by looking at it with your eye, and that'll make
it much easier in post, because any corrections that you
have to make will be minor ones. As a fourth option, if you don't
think you're gonna remember to set your white balance,
or you just don't think you'll have time to,
I would suggest setting a manual
white balance the same way by adjusting the kelvin
to a neutral setting between, like, 4500K and 6000K,
and just leave it there. That way, if you do need
to make adjustments, it's not gonna be too extreme
because you're somewhat in the middle, and also
you'll have the advantage of having consistency
throughout your shots. Which is why I put
auto white balance as my fifth
and last place option. As you saw in my first example, auto white balance
isn't the most reliable. It's a lot better
than being off by 2000K, but the problem
is how it changes. If you shoot using sunlight, whether outdoors
or near a window, the temperature
can change constantly when using
auto white balance. So that's why I'd rather
set it to something manual, like 5500K, and then even
if I need to adjust it in post, at least I have
something consistent rather than shifting
throughout the shoot. Some cameras allow you
to lock the auto white balance, which is better, because that
way when you let the camera set the white balance
automatically, you can lock it in place
so that when clouds shift in front of the sun,
it's not gonna change the white balance
as you go through your shoot. Now this method's okay,
but to me it kind of reinforces the idea that you never
actually have to learn proper techniques
for white balancing, and I think that's a bad thing. And also,
like we were talking about, cameras aren't perfect
at white balance, whether it's that their
metering is a little bit off or that they're
sensitive to certain tints, so I find that
it's never quite as accurate as if you set it yourself, and it's not consistent
between brands. However, auto white balance
is usually fine if your images aren't gonna be juxtaposed,
because then you won't be able to see the inconsistencies
in white balance like you would if they were
in a sequence or collection. But I would not recommend
this for video, because then every time
you cut and resume, you're gonna have
a different white balance. And speaking of video, let's take a look
at the video samples. So here we've got some
footage shot in 8-bit H.264 with a standard profile,
and at the same three manual custom
white balance temperatures from before-- the 5500,
the 2500, and the 10,000. So let's just grab
the dropper here and put it in the same corner. Similar in results to what
we got in the photos, you know, added a little bit of magenta,
and the result is fine, the colours look
OK on the Rubik's Cube. Now let's jump over
to the 10,000K one and dropper it
in the same spot... and again, you know,
it looks OK. We have some of the same issues
that we had with JPEG, where we can see that
our oranges are more subdued when compared to the original,
and we have this other issue that
I want to tell you about which involves the highlights.
Remember, I said if things are overexposed,
the problems get worse. And sometimes this happens
and you don't realize it, or there might be something
that wasn't part of the subject but it was in the background,
that when the white balance was OK you didn't mind that
it was overexposed, but now when you change
the white balance, you're like, “that is disgusting,”
and that's what happens here. Look at this crazy,
yellow, blown out, glowy gradient that
gets added with, like, maybe even a little bit
of green in it. And it's not something
that can be easily fixed. Like, if even if you
bring the highlights down, now all you're left with is
this really weird yellow thing. So you're gonna have to go
in and be tweaking all kinds of stuff to try
and restore that skin, and if you've done this before,
it still never looks great, and it takes so much work,
and you just hate yourself. Like “why didn't I just
set the white balance?” And now if we do the 2500K and
dropper it in the top corner... [laughs]
It's just absolutely terrible. Uh-- [chuckles] Do I even
need to say anything about it? It’s-- it’s awful, right? So as you can see,
unless you record in a proper raw format,
videos are subject to the same limitations when it comes
to white balance as JPEGs are. So consider them
just like moving JPEGs. And in the case of some
cameras’ codecs, that's exactly what they are,
actually- moving JPEGs. So what about
10-bit Log footage? Well, over here I have
three shots that are all... shot in 10-bit,
and I've already applied a LUT to conform them to Rec.709 and made sure that
the exposure was even. So this isn't
an overexposed/unexposed, this is more
of a neutral exposure. So after droppering it,
very small adjustments. It warmed it up by,
like, one notch, and by a magenta by 0.8, and I think that it looks
like a pretty decent shot. You know, we can increase
the exposure if we want, we can increase
the saturation of it a bit, and it's fine in both cases. Now let's try
the 10,000K image. And we'll dropper it
up here... you know, again, not too bad. If we compare
that to the 5500K, I would say that
it did reasonably well. There is a bit-- on the skin,
there is a bit more... ...just,
it doesn't blend as well. I find that-- like,
this one has, you know, natural roll-offs,
and on this one, there are a bit of areas where
the colour's kind of blotchy. You can see some here
at my fingertip that isn't really as problematic
on the 5500K shot. But overall,
I would say this is decent, and I would say that
this result is better than the 8-bit
standard profile version. And that could be
because of the 10-bit, it could also be
because of the Log. But once again,
if we try the 2500K version and we dropper it,
it doesn't even matter that it's 10-bit,
it doesn't matter that it's Log, you just can't
shift the white balance that hard in any kind of video
unless it's proper raw video. So in conclusion,
raw photos and raw video-- yes, you can dramatically
change the white balance in post without
really losing anything. But everything else--
JPEGs, 8-bit video, 10-bit video--
I shot these with DNx, so it doesn't
really matter the bit rate. You can give it lots of bits,
Log, not Log-- you can't do it. You just can't shift
the white balance that much. Small adjustments, yes.
Large adjustments, no. And so these tests emphasize
how important it is to put thought into your white
balance every time you shoot. And I think this
is a good practice to keep, even if you shoot raw. But that's gonna be it for me. I hope you found
this video helpful, or at least entertaining. 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, feel free to hit
the dislike button twice. Alright...
I'm done. ♪
