Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:12,866 --> 00:00:15,166
hi in our first lesson
2
00:00:15,166 --> 00:00:18,233
I will be reviewing the basics of deep compositing
3
00:00:18,766 --> 00:00:21,199
for viewers with no previous deep experience
4
00:00:24,066 --> 00:00:27,866
here I have a simple deep image
5
00:00:29,333 --> 00:00:32,999
and this image is composed of two different parts
6
00:00:33,400 --> 00:00:34,866
it's composed of
7
00:00:36,933 --> 00:00:39,266
the motion blurred red sphere element
8
00:00:39,500 --> 00:00:42,033
as well as the blue cylinders element
9
00:00:43,466 --> 00:00:46,066
now you might be asking
10
00:00:46,666 --> 00:00:49,433
what is deep compositing and how does it work
11
00:00:49,900 --> 00:00:53,600
so in a normal CG render
12
00:00:55,200 --> 00:00:59,033
every pixel in the image gets one sample
13
00:00:59,500 --> 00:01:00,933
one sample of color
14
00:01:00,933 --> 00:01:02,333
one sample of depth
15
00:01:02,333 --> 00:01:05,366
one sample of every other AOV
16
00:01:05,733 --> 00:01:09,766
recorded in the output file for that pixel
17
00:01:12,466 --> 00:01:14,799
the difference is with deep
18
00:01:15,400 --> 00:01:18,400
you get multiple samples per pixel
19
00:01:21,100 --> 00:01:22,500
so what happens is
20
00:01:23,700 --> 00:01:29,233
the CG scene is divided up into multiple depth slices
21
00:01:30,500 --> 00:01:33,366
and each slice gets its own sample recorded
22
00:01:35,300 --> 00:01:36,633
and I can show you this
23
00:01:38,500 --> 00:01:40,100
using the deep sample note
24
00:01:43,566 --> 00:01:44,533
as I search around
25
00:01:44,533 --> 00:01:46,033
you can see that
26
00:01:46,700 --> 00:01:49,200
every pixel has a different set of samples
27
00:01:51,100 --> 00:01:54,566
now it's important to understand that
28
00:01:55,666 --> 00:01:57,933
deep sampling is adaptive
29
00:01:57,933 --> 00:01:59,266
it's not constant
30
00:01:59,800 --> 00:02:01,733
so in situations where
31
00:02:01,733 --> 00:02:02,766
for instance
32
00:02:04,533 --> 00:02:06,433
we have one sample per pixel
33
00:02:09,366 --> 00:02:12,466
that's because there is nothing behind this pixel
34
00:02:12,766 --> 00:02:15,633
there is no other cylinder behind this
35
00:02:16,366 --> 00:02:19,233
so only one sample is required
36
00:02:20,066 --> 00:02:23,566
to accurately describe the depth of this pixel
37
00:02:23,866 --> 00:02:25,199
when we get into
38
00:02:26,500 --> 00:02:28,933
layering where there's multiple cylinders
39
00:02:28,933 --> 00:02:30,499
then we get more samples
40
00:02:30,866 --> 00:02:33,233
here at this edge between
41
00:02:34,166 --> 00:02:37,166
the foreground cylinder and the midground cylinder
42
00:02:37,466 --> 00:02:38,833
we see there's two samples
43
00:02:40,166 --> 00:02:42,899
we get into things like the sphere
44
00:02:43,266 --> 00:02:46,099
we get many more samples because it's motion blurt
45
00:02:46,100 --> 00:02:48,300
has a lot of transparency or opacity
46
00:02:48,600 --> 00:02:51,766
and it requires a lot more samples
47
00:02:52,500 --> 00:02:55,000
to accurately describe its depth
48
00:02:55,733 --> 00:02:59,666
here we get 106 samples at the single pixel
49
00:03:03,200 --> 00:03:06,166
so I can visualize this for you a different way
50
00:03:07,400 --> 00:03:09,833
I can visualize it in Nuke's 3D space
51
00:03:09,933 --> 00:03:12,499
using the deep 2 points node
52
00:03:12,500 --> 00:03:14,033
I can generate a point cloud
53
00:03:19,300 --> 00:03:20,400
so here we see
54
00:03:23,600 --> 00:03:24,700
a 3D point cloud
55
00:03:24,700 --> 00:03:28,866
reconstruction of the 3D scene from Maya
56
00:03:29,766 --> 00:03:34,433
and we can see that the cylinders behind this sphere
57
00:03:34,566 --> 00:03:36,099
this is the camera over here
58
00:03:36,100 --> 00:03:37,966
so from the camera's perspective
59
00:03:39,866 --> 00:03:41,566
which is about right here
60
00:03:43,066 --> 00:03:46,666
there is still data for these cylinders
61
00:03:48,500 --> 00:03:50,800
now when we compare that to
62
00:03:51,066 --> 00:03:54,466
the kind of data that gets recorded in something like
63
00:03:56,766 --> 00:03:58,033
a depth pass
64
00:03:59,100 --> 00:04:01,833
we can see where there's only one sample per pixel
65
00:04:03,133 --> 00:04:06,366
where the sphere excludes the cylinder
66
00:04:06,966 --> 00:04:09,633
there's no more information it's gone
67
00:04:13,566 --> 00:04:15,399
so what does this allow us to do
68
00:04:19,133 --> 00:04:21,066
but we can do really interesting things
69
00:04:23,766 --> 00:04:24,466
what we can do
70
00:04:24,466 --> 00:04:25,199
for example
71
00:04:25,200 --> 00:04:30,600
is we can actually move the objects around
72
00:04:30,600 --> 00:04:32,266
normally this isn't possible
73
00:04:32,266 --> 00:04:34,399
and it's not possible because
74
00:04:35,800 --> 00:04:38,066
in a normal workflow
75
00:04:38,300 --> 00:04:39,766
with one sample per pixel
76
00:04:39,766 --> 00:04:43,566
using a beauty Pass and an alpha channel
77
00:04:43,566 --> 00:04:48,199
and a depth pass and all the other traditional AOVs
78
00:04:48,600 --> 00:04:50,466
you need to render holdouts
79
00:04:51,700 --> 00:04:54,800
so if this fear is animated
80
00:04:55,266 --> 00:04:57,466
it would go through multiple iterations
81
00:04:57,733 --> 00:04:59,066
you'd have to re render
82
00:04:59,066 --> 00:05:02,399
and every time you re render with a new animation
83
00:05:03,266 --> 00:05:05,333
you would have to render holdouts because
84
00:05:05,333 --> 00:05:07,899
if all you have for compositing is an alpha channel
85
00:05:09,000 --> 00:05:11,300
you need to have alpha channels that don't overlap
86
00:05:13,800 --> 00:05:15,033
so it means that
87
00:05:15,800 --> 00:05:17,166
for these pixels
88
00:05:17,733 --> 00:05:19,566
the alpha channel of the cylinders
89
00:05:19,566 --> 00:05:22,066
would need to become progressively
90
00:05:22,400 --> 00:05:24,333
more and more transparent
91
00:05:24,333 --> 00:05:26,266
and then eventually disappear completely
92
00:05:26,533 --> 00:05:29,599
and the inverse would happen to correspond
93
00:05:29,600 --> 00:05:32,166
for transparency in the sphere element
94
00:05:33,333 --> 00:05:39,166
so you would have holdouts for the animation
95
00:05:39,166 --> 00:05:41,966
baked into the CG renders with deep
96
00:05:42,766 --> 00:05:43,833
the holdouts
97
00:05:45,066 --> 00:05:46,499
are generated in compositing
98
00:05:46,500 --> 00:05:50,300
that holdout process is deferred to compositing
99
00:05:51,400 --> 00:05:54,833
so the CG render is rendered as a whole
100
00:05:57,133 --> 00:06:00,066
see the cylinders are whole there's no holdouts
101
00:06:00,766 --> 00:06:02,366
the sphere no holdouts
102
00:06:05,666 --> 00:06:08,166
but when I combine them together
103
00:06:09,933 --> 00:06:11,799
they're accurately merged
104
00:06:11,866 --> 00:06:13,566
that's because of the deep data
105
00:06:13,800 --> 00:06:18,000
the deep data is describing how they should intersect
106
00:06:20,200 --> 00:06:22,400
now something interesting about this is that
107
00:06:22,400 --> 00:06:24,400
because we have a complete
108
00:06:25,266 --> 00:06:28,299
volumetric sample of these two different elements
109
00:06:28,566 --> 00:06:31,066
I can actually change their relationship
110
00:06:31,200 --> 00:06:32,500
and accurately
111
00:06:33,700 --> 00:06:36,466
reconstruct what the elements would look like
112
00:06:36,466 --> 00:06:38,033
in a different relationship
113
00:06:38,466 --> 00:06:39,833
so for example
114
00:06:43,800 --> 00:06:47,900
I can put this behind all the cylinders
115
00:06:47,900 --> 00:06:49,200
I can push it back
116
00:06:49,966 --> 00:06:53,966
and you can see as I as it moves through through space
117
00:06:56,700 --> 00:07:00,666
it's as if I'm natively rendering this
118
00:07:01,700 --> 00:07:05,700
it looks like it would if it was just rendered this way
119
00:07:07,866 --> 00:07:08,599
zoom in a little bit
120
00:07:08,600 --> 00:07:12,800
so we have a better view of this as we go through
121
00:07:14,533 --> 00:07:16,433
this normally isn't possible
122
00:07:18,400 --> 00:07:22,333
but we can do these kinds of things in compositing
123
00:07:22,333 --> 00:07:25,233
because of the deep samples
124
00:07:29,600 --> 00:07:32,366
now at this point you might be thinking
125
00:07:32,366 --> 00:07:34,199
I know what a depth pass looks like
126
00:07:34,200 --> 00:07:36,133
I know what a position pass looks like
127
00:07:36,133 --> 00:07:38,666
but what is a deep pass look like
128
00:07:39,400 --> 00:07:40,666
so if we go to the viewer
129
00:07:42,166 --> 00:07:43,566
who go to the deep layer
130
00:07:44,700 --> 00:07:45,300
can see here
131
00:07:45,300 --> 00:07:46,400
this is deep
132
00:07:47,566 --> 00:07:48,966
so why is it yellow
133
00:07:49,466 --> 00:07:53,366
it's yellow because it only has two channels
134
00:07:53,766 --> 00:07:55,499
red and green
135
00:07:55,933 --> 00:07:58,233
and the values in this particular case
136
00:07:59,000 --> 00:08:00,333
are all super white
137
00:08:00,333 --> 00:08:01,866
they're all above 1
138
00:08:02,800 --> 00:08:03,566
so what do we see
139
00:08:03,566 --> 00:08:06,599
we see a clipped red and green channel
140
00:08:06,600 --> 00:08:08,466
which gives us yellow
141
00:08:08,466 --> 00:08:09,466
solid yellow
142
00:08:12,100 --> 00:08:12,900
now
143
00:08:14,166 --> 00:08:15,499
as I move my cursor around
144
00:08:15,500 --> 00:08:18,133
I can see the values are changing
145
00:08:18,133 --> 00:08:19,599
and of course we know the values change in deep
146
00:08:19,600 --> 00:08:21,300
because we use the deep sample
147
00:08:22,900 --> 00:08:23,800
can see here
148
00:08:23,866 --> 00:08:24,466
move around
149
00:08:24,466 --> 00:08:25,666
we see the values
150
00:08:26,933 --> 00:08:29,833
over here the values change on every pixel
151
00:08:32,266 --> 00:08:33,233
so
152
00:08:34,766 --> 00:08:37,066
if I expose down
153
00:08:37,366 --> 00:08:40,199
in this case to F2 56
154
00:08:41,366 --> 00:08:43,299
we see something that looks more familiar
155
00:08:43,500 --> 00:08:45,300
we see the cylinders
156
00:08:45,300 --> 00:08:46,500
we see the sphere
157
00:08:47,166 --> 00:08:49,233
and this is just raw deep data
158
00:08:49,366 --> 00:08:50,766
and there's no other way to really view it
159
00:08:50,766 --> 00:08:54,966
in the viewer as a 2D thing it's just
160
00:08:54,966 --> 00:08:57,433
it's just what it is and it's a
161
00:08:57,533 --> 00:08:58,899
it's a technical pass
162
00:08:59,900 --> 00:09:01,766
it's not really meant to be looked at
163
00:09:01,866 --> 00:09:04,566
and at no point do you really ever need to look at it
164
00:09:04,566 --> 00:09:07,166
there's nothing you really learn from looking at it
165
00:09:07,700 --> 00:09:08,866
I just want to show it to you
166
00:09:08,866 --> 00:09:12,599
so you kind of understand how nuke interprets it
167
00:09:12,600 --> 00:09:15,400
but deep itself as a
168
00:09:15,400 --> 00:09:16,666
as a visual thing is
169
00:09:16,666 --> 00:09:18,466
is sort of abstract
170
00:09:19,366 --> 00:09:21,899
now two channels
171
00:09:22,800 --> 00:09:26,633
I mentioned that there was multiple samples per pixel
172
00:09:27,700 --> 00:09:29,533
but why are there two channels
173
00:09:29,533 --> 00:09:31,899
the reason is because every sample
174
00:09:32,866 --> 00:09:35,866
per slice actually has a front and a back
175
00:09:37,766 --> 00:09:40,499
so as we go through here we see at this Pixel
176
00:09:41,900 --> 00:09:43,500
there are front and back
177
00:09:43,900 --> 00:09:47,400
there are front and back slices so
178
00:09:49,266 --> 00:09:50,633
can see all 32
179
00:09:52,333 --> 00:09:53,499
and if we look at the difference here
180
00:09:53,500 --> 00:09:55,766
we look at the red channel this is the front
181
00:09:55,766 --> 00:09:58,099
look at the green channel this is the back
182
00:10:02,666 --> 00:10:04,466
so that's what deep looks like
183
00:10:05,400 --> 00:10:07,100
that's why it's got two channels
184
00:10:08,366 --> 00:10:11,433
the last and one of the most important things to cover
185
00:10:12,733 --> 00:10:15,399
is deep nodes
186
00:10:15,400 --> 00:10:18,366
now deep data is
187
00:10:19,533 --> 00:10:22,633
a kind of data that's only compatible with deep nodes
188
00:10:22,900 --> 00:10:27,266
you can't use regular nuke nodes with deep data
189
00:10:28,500 --> 00:10:31,666
you can think about deep as something like
190
00:10:31,666 --> 00:10:35,099
the way you would use Nuke's 3D nodes
191
00:10:35,200 --> 00:10:36,966
or Nuke's particle nodes
192
00:10:37,133 --> 00:10:38,033
they're not
193
00:10:38,166 --> 00:10:41,599
you can't just plug in a regular 2D node at any place
194
00:10:41,600 --> 00:10:43,266
anytime and expect it to work
195
00:10:43,266 --> 00:10:44,066
it's because
196
00:10:44,400 --> 00:10:46,866
the kind of data that's flowing through the pipe
197
00:10:46,933 --> 00:10:50,033
is not the kind of data that the other nodes
198
00:10:50,566 --> 00:10:51,633
can process
199
00:10:51,933 --> 00:10:52,699
so for instance
200
00:10:52,700 --> 00:10:54,666
let's say I wanted to take this image
201
00:10:54,900 --> 00:10:55,766
and blur it
202
00:10:55,766 --> 00:10:56,666
create a blur node
203
00:10:56,666 --> 00:10:57,699
try to plug it in
204
00:10:57,966 --> 00:10:59,199
it won't connect
205
00:11:01,766 --> 00:11:03,466
there are two nodes that
206
00:11:03,466 --> 00:11:06,233
you would use on a regular basis that will connect
207
00:11:07,533 --> 00:11:09,599
and they are the dot
208
00:11:12,133 --> 00:11:13,299
and the no op
209
00:11:14,333 --> 00:11:16,999
the reason these two nodes work with deep
210
00:11:17,066 --> 00:11:19,066
is because they're just pass throughs
211
00:11:19,300 --> 00:11:22,833
they don't perform any processing or functions at all
212
00:11:23,333 --> 00:11:27,999
they're really just visual aids on our node graph
213
00:11:30,200 --> 00:11:35,233
so I'll show you in later courses
214
00:11:38,400 --> 00:11:40,100
how to deal with deep nodes
215
00:11:40,100 --> 00:11:41,866
how to deal with this restriction
216
00:11:42,100 --> 00:11:44,033
and it's a pretty severe restriction
217
00:11:44,733 --> 00:11:47,666
is really a very limited number of deep nodes
218
00:11:48,666 --> 00:11:50,766
and I'll show you when
219
00:11:51,933 --> 00:11:53,766
it's advantageous to use deep nodes
220
00:11:53,766 --> 00:11:57,433
and when it's more advantageous to convert to
221
00:11:57,900 --> 00:12:01,866
a regular 2D image and work with your comp that way
222
00:12:02,166 --> 00:12:03,633
so I hope you stick with me
223
00:12:04,500 --> 00:12:05,866
and I'll see you in the next course
15886
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.