WEBVTT
Kind: captions
Language: en

1
00:00:00.000 --> 00:00:03.993
This video is sponsored by CuriosityStream,
and my own Nebula Original series.

2
00:00:03.993 --> 00:00:04.993
More on that later.

3
00:00:04.993 --> 00:00:07.920
I’ve been using a green-screen for my
recent computer-science videos.

4
00:00:07.920 --> 00:00:13.120
I don’t like using green-screens, I think there’s
no substitute for filming in a real physical location,

5
00:00:13.120 --> 00:00:15.460
but I haven't had much choice lately.

6
00:00:15.460 --> 00:00:18.316
And some folks, very kindly,
said they couldn’t tell,

7
00:00:18.316 --> 00:00:21.738
Now, a lot of that is down to my animator
for these videos,

8
00:00:21.738 --> 00:00:24.300
but a lot of that is down to
some good planning.

9
00:00:24.300 --> 00:00:27.730
Big-budget television shows get
green-screen wrong. All the time.

10
00:00:27.730 --> 00:00:30.380
This clip from 30 Rock looks… dreadful,

11
00:00:30.380 --> 00:00:33.030
but it’s difficult to immediately
point to why.

12
00:00:33.030 --> 00:00:35.291
Well, first of all, there is the key.

13
00:00:35.291 --> 00:00:39.010
That’s the term for how the computer works out
what’s background and foreground.

14
00:00:39.010 --> 00:00:41.340
There are all sorts of settings you can use
to tweak this:

15
00:00:41.340 --> 00:00:43.411
the exact colour of green it’s looking for,

16
00:00:43.411 --> 00:00:45.390
how close to that green it has to be,

17
00:00:45.390 --> 00:00:49.780
how it deals with colour in the transition
between background and foreground,

18
00:00:49.780 --> 00:00:52.410
how much it smooths out
fine details like hair.

19
00:00:52.410 --> 00:00:57.079
Get it wrong, and hair can appear and disappear
as someone turns their head.

20
00:00:57.079 --> 00:00:59.489
But let’s assume that you get a perfect key.

21
00:00:59.489 --> 00:01:01.590
It can still look weird.

22
00:01:01.590 --> 00:01:04.519
And that’s because if you want convincing
green-screen footage,

23
00:01:04.519 --> 00:01:09.719
you have to film with a camera that more-or-less
matches the one that filmed the background.

24
00:01:09.719 --> 00:01:13.139
Even for all the technical stuff that
most people have no idea about:

25
00:01:13.139 --> 00:01:16.920
because even if your viewers have no clue
why it looks weird,

26
00:01:16.920 --> 00:01:19.359
they’ll still be able to tell something’s wrong.

27
00:01:19.359 --> 00:01:22.570
Before that, though, you have to
match the lighting.

28
00:01:22.570 --> 00:01:24.479
You’d think this one’s obvious, right?

29
00:01:24.479 --> 00:01:27.829
But have a look at this scene from
The Big Bang Theory,

30
00:01:27.829 --> 00:01:32.819
who despite being one of the most popular
shows in America when this episode aired,

31
00:01:32.819 --> 00:01:37.200
apparently couldn’t afford to film at some
rocks twenty miles outside Los Angeles.

32
00:01:37.200 --> 00:01:40.009
Sure, you could argue it’s a reference to
Star Trek

33
00:01:40.009 --> 00:01:45.889
who used to have alien planets filmed on stages
with some rocks unconvincingly strewn around, but…

34
00:01:45.889 --> 00:01:46.889
I’m not convinced.

35
00:01:46.889 --> 00:01:49.319
Anyway: the sun is clearly over there.

36
00:01:49.319 --> 00:01:51.479
The rocks are in harsh shadow,
there’s haze in the background.

37
00:01:51.479 --> 00:01:55.060
And the actors are…
lit like they’re on a stage.

38
00:01:55.060 --> 00:01:59.479
This guy’s head casts a shadow on this guy’s
chest in completely the wrong direction.

39
00:01:59.479 --> 00:02:03.270
Even if you nail the lighting, you’ve got
to get the colour balance right.

40
00:02:03.270 --> 00:02:07.560
Daylight has more blue in it, most indoor
lights have more orange in them.

41
00:02:07.560 --> 00:02:10.469
Screw that up, and the
different locations are clear.

42
00:02:10.469 --> 00:02:14.239
Next, you’ve got to match the shutter speed.

43
00:02:14.239 --> 00:02:16.739
So in this shot, filmed outside
under bright sunlight,

44
00:02:16.739 --> 00:02:18.870
you can see there’s almost no motion blur,

45
00:02:18.870 --> 00:02:20.950
even on fast-moving vehicles
that are passing by.

46
00:02:20.950 --> 00:02:24.260
The shutter speed is short:
in a thousandth of a second,

47
00:02:24.260 --> 00:02:27.680
the camera sensor can get all the light it
needs for a clear picture.

48
00:02:27.680 --> 00:02:31.019
But I’m indoors right now. I can’t
match how bright that sunlight is,

49
00:02:31.019 --> 00:02:36.260
so the camera sensor has to sample for about
ten times as long to get a decent picture,

50
00:02:36.260 --> 00:02:39.100
which is enough time for me
to blur when I move.

51
00:02:39.100 --> 00:02:42.860
And if I’m blurred, then the background isn’t,

52
00:02:42.860 --> 00:02:44.370
and then it’s going to look wrong.

53
00:02:44.370 --> 00:02:47.220
And that blur is also really difficult to
key out as well,

54
00:02:47.220 --> 00:02:49.290
but that’s a separate problem.

55
00:02:49.290 --> 00:02:50.780
Next up: framing.

56
00:02:50.780 --> 00:02:53.010
If the person on screen appears to be hovering
in mid-air,

57
00:02:53.010 --> 00:02:56.859
or is shot straight-on while the camera is
looking up or down at the background,

58
00:02:56.859 --> 00:02:59.030
obviously it’s going to look wrong.

59
00:02:59.030 --> 00:03:00.819
But there are more subtle things too.

60
00:03:00.819 --> 00:03:02.800
Now, I was really lucky here,

61
00:03:02.800 --> 00:03:08.127
I remember how the camera was set up for this
original background footage in the computer museum.

62
00:03:08.127 --> 00:03:10.840
It was four or five metres away,
zoomed in quite a bit,

63
00:03:10.840 --> 00:03:14.260
with the camera positioned just above my eyeline,
so I’m looking up a little.

64
00:03:14.260 --> 00:03:17.849
Means I don’t end up with too much of a
double chin.

65
00:03:17.849 --> 00:03:20.089
But let’s just pause for a moment --

66
00:03:20.089 --> 00:03:21.849
-- okay.

67
00:03:21.849 --> 00:03:24.560
The camera is now physically closer
but zoomed out.

68
00:03:24.560 --> 00:03:26.880
And suddenly, this looks wrong.

69
00:03:26.880 --> 00:03:28.390
The same parts of my body are in the shot,

70
00:03:28.390 --> 00:03:29.470
in roughly the same positions,

71
00:03:29.470 --> 00:03:32.069
but I don’t fit into the background nearly as well.

72
00:03:32.069 --> 00:03:35.053
This is not how I would look sitting
in that chair in the real Museum.

73
00:03:35.053 --> 00:03:37.916
I couldn't reach out and nearly touch the camera.

74
00:03:37.916 --> 00:03:40.370
People’s faces look different,
everything looks different,

75
00:03:40.370 --> 00:03:42.019
depending on how close the camera is.

76
00:03:42.019 --> 00:03:44.400
In this shot from
Guardians of the Galaxy 2,

77
00:03:44.400 --> 00:03:47.450
the camera is changing distance and zoom at
the same time,

78
00:03:47.450 --> 00:03:52.980
and the difference in how the actor’s face
looks at the start and end is remarkable.

79
00:03:52.980 --> 00:03:55.041
Okay, let’s go back.

80
00:03:55.041 --> 00:03:58.540
Okay, that’s better!
If you want your green screen to be convincing,

81
00:03:58.540 --> 00:04:01.519
those are just some of the things
you’ve got to match.

82
00:04:01.519 --> 00:04:03.620
There may be others that I’ve missed, too.

83
00:04:03.620 --> 00:04:06.989
But there is another approach,
and I’d say it’s a much better idea.

84
00:04:06.989 --> 00:04:08.689
Don’t try to be convincing.

85
00:04:08.689 --> 00:04:11.340
For those of us who have the creative option,

86
00:04:11.340 --> 00:04:14.830
it’s a lot easier to just acknowledge the
green screen, lean into it,

87
00:04:14.830 --> 00:04:19.230
and just put something interesting up there
instead that isn’t meant to be real.

88
00:04:19.230 --> 00:04:23.620
It’s not as good as filming in an actual
location, but at least it’s honest.

89
00:04:23.620 --> 00:04:27.210
I've got an original series over on Nebula.
Here's the trailer.

90
00:04:27.210 --> 00:04:29.800
"I invited five people to play some games."

91
00:04:29.800 --> 00:04:31.800
- I trust no-one.
- None of us are trustworthy.

92
00:04:31.800 --> 00:04:35.309
"In an environment designed to slowly break
their team apart."

93
00:04:35.309 --> 00:04:36.470
- This is real money.

94
00:04:36.470 --> 00:04:38.849
"But all they knew is they'd be sat round
a table

95
00:04:38.849 --> 00:04:42.120
"trying to win real cash. $10,000."

96
00:04:42.120 --> 00:04:44.270
- The vibe's changed after that theft.

97
00:04:44.270 --> 00:04:47.340
"This is a show about trust, about loyalty,

98
00:04:47.340 --> 00:04:48.880
"and about money."

99
00:04:48.880 --> 00:04:50.566
- Tom wants the chaos.

100
00:04:52.050 --> 00:04:56.460
Along with my series, Nebula has Wendover
Productions' original hour-long documentary,

101
00:04:56.460 --> 00:04:58.340
The Final Years of Majuro,

102
00:04:58.340 --> 00:05:00.340
and loads of other content and collaborations

103
00:05:00.340 --> 00:05:03.720
from education video creators
that you may well have heard of.

104
00:05:03.720 --> 00:05:07.215
It's bundled at this link with CuriosityStream,
a subscription streaming service with

105
00:05:07.215 --> 00:05:11.229
thousands of big-budget, professional
documentaries and nonfiction titles.

106
00:05:11.229 --> 00:05:16.750
CuriosityStream is $2.99/month or only $19.99/year,
including access to Nebula and my series,

107
00:05:16.750 --> 00:05:20.160
when you sign up at curiositystream.com/tomscott.