Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:18,400 --> 00:00:19,333
in this lesson
2
00:00:19,333 --> 00:00:21,366
we'll discuss the origins and history
3
00:00:21,366 --> 00:00:22,833
of deep compositing
4
00:00:23,866 --> 00:00:27,199
so the story of deep compositing begins in 2000
5
00:00:27,200 --> 00:00:29,066
at Pixar Animation Studios
6
00:00:29,800 --> 00:00:34,866
so Pixar was searching for a solution to rendering
7
00:00:35,500 --> 00:00:41,000
self shadows and cash shadows for hair for
8
00:00:41,133 --> 00:00:42,699
and volumetric renders
9
00:00:44,533 --> 00:00:48,866
now this is a Cigraph paper that Pixar published
10
00:00:48,933 --> 00:00:52,233
describing their new Deep Shadow Map technology
11
00:00:54,000 --> 00:00:57,233
so normally
12
00:00:58,766 --> 00:01:00,366
what you would have to do
13
00:01:00,700 --> 00:01:05,166
if you rendered something like a fur or hair object
14
00:01:05,166 --> 00:01:08,266
this is a example as a ball with 50,000 hairs
15
00:01:10,166 --> 00:01:13,433
is that you'd have to use a shadow map
16
00:01:13,566 --> 00:01:16,499
now shadow map has a finite number of pixels
17
00:01:17,800 --> 00:01:23,200
and those pixels have a memory footprint
18
00:01:24,300 --> 00:01:26,466
so if you look at something that you might
19
00:01:26,733 --> 00:01:28,333
find more desirable in production
20
00:01:28,333 --> 00:01:31,366
something like a 5 12 by 5 12 pixel shadow map
21
00:01:31,566 --> 00:01:35,166
you see the result is coarse and grainy
22
00:01:35,166 --> 00:01:37,999
and not really satisfactory for production
23
00:01:38,500 --> 00:01:40,266
so you increase your samples
24
00:01:41,100 --> 00:01:44,000
to something like a 4K by 4K shadow map
25
00:01:44,933 --> 00:01:49,999
and you get a satisfactory image
26
00:01:50,266 --> 00:01:52,566
but the memory hit
27
00:01:52,566 --> 00:01:55,366
and the computational hit is unsatisfactory
28
00:01:56,000 --> 00:01:57,366
it's very large
29
00:01:57,700 --> 00:01:59,666
and if you're doing this all the time
30
00:02:00,066 --> 00:02:01,966
it's going to impact your production
31
00:02:03,300 --> 00:02:04,200
so
32
00:02:06,200 --> 00:02:07,533
this fourth image
33
00:02:07,533 --> 00:02:10,433
you see a 5 tail by 5 tail deep shadow map
34
00:02:11,333 --> 00:02:12,566
now the Deep Shadow map
35
00:02:12,766 --> 00:02:14,666
unlike a normal shadow map of course
36
00:02:14,866 --> 00:02:18,799
has deep slices so it has multiple samples
37
00:02:20,333 --> 00:02:21,133
and
38
00:02:22,933 --> 00:02:26,199
this allows you to get the same self shadowing
39
00:02:26,200 --> 00:02:29,033
and cash shadowing quality
40
00:02:29,366 --> 00:02:30,699
but with a smaller map
41
00:02:32,300 --> 00:02:33,266
so
42
00:02:35,266 --> 00:02:36,599
what Pixar did
43
00:02:39,333 --> 00:02:41,866
he has developed the technology that ended up
44
00:02:42,133 --> 00:02:43,999
becoming deep compositing
45
00:02:44,266 --> 00:02:47,799
to solve a somewhat related issue
46
00:02:47,800 --> 00:02:49,633
because when you think about
47
00:02:51,366 --> 00:02:53,666
the same reasons that we use deep compositing
48
00:02:53,666 --> 00:02:53,999
it's really
49
00:02:54,000 --> 00:02:56,800
the same reasons that Pixar uses deep shadow maps
50
00:02:57,800 --> 00:03:01,800
they use it for dealing with motion blur
51
00:03:01,800 --> 00:03:03,200
dealing with Volumetrics
52
00:03:03,200 --> 00:03:06,800
and dealing with complex occlusions
53
00:03:08,766 --> 00:03:11,599
so at the end of this paper here
54
00:03:11,600 --> 00:03:13,433
you see they give examples of
55
00:03:14,066 --> 00:03:17,599
a shadow map for a heavily motion blurred object
56
00:03:17,966 --> 00:03:21,466
self shadowing and volumetric shadows
57
00:03:22,733 --> 00:03:27,966
so these are really the same examples of
58
00:03:27,966 --> 00:03:32,366
of the types of renders that we would use to
59
00:03:32,900 --> 00:03:34,466
composite with deep anyway
60
00:03:34,466 --> 00:03:36,766
so can see that it's really
61
00:03:36,766 --> 00:03:38,966
the same problem they're trying to solve
62
00:03:39,666 --> 00:03:42,599
and this is something that
63
00:03:43,066 --> 00:03:44,299
that comes into play
64
00:03:44,300 --> 00:03:48,666
and people in the compositing world recognize this
65
00:03:50,133 --> 00:03:51,166
so
66
00:03:52,766 --> 00:03:55,133
Pixar after they publish this paper
67
00:03:55,133 --> 00:03:56,633
they go ahead and they implement
68
00:03:57,133 --> 00:03:59,766
deep shadow maps into their render man software
69
00:04:01,300 --> 00:04:04,566
so deep shadow maps become commercially available
70
00:04:04,700 --> 00:04:06,433
and it uses the format
71
00:04:07,966 --> 00:04:11,466
Detex it's a deep shadow a deep deep texture
72
00:04:12,733 --> 00:04:17,233
and this is the format that Render Man still uses today
73
00:04:19,366 --> 00:04:20,299
so
74
00:04:22,500 --> 00:04:24,300
we jumped to what
75
00:04:24,300 --> 00:04:26,433
a digital in 2,008
76
00:04:27,166 --> 00:04:29,299
and a man named Calm Doncaster
77
00:04:30,400 --> 00:04:34,633
he recognizes what Pixar has done with deep chetamaps
78
00:04:34,766 --> 00:04:38,433
and he thinks that that kind of
79
00:04:40,166 --> 00:04:43,199
depth slicing from a 3D scene
80
00:04:43,366 --> 00:04:47,566
would be a great way to control holdouts in compositing
81
00:04:48,533 --> 00:04:50,299
so he goes and he builds
82
00:04:51,000 --> 00:04:54,266
an implementation of deep compositing in shake
83
00:04:54,266 --> 00:04:55,866
so the time whether was still shake
84
00:04:55,866 --> 00:04:56,966
this would be for nuke
85
00:04:58,700 --> 00:05:00,800
and they use it on the scene
86
00:05:00,966 --> 00:05:03,366
in the day of your stood still
87
00:05:03,666 --> 00:05:07,366
where the truck is dissolved
88
00:05:07,733 --> 00:05:09,099
and it's a volumetric render
89
00:05:09,100 --> 00:05:12,000
and it's a really good example of where you'd use deep
90
00:05:12,800 --> 00:05:15,433
so this was a successful test
91
00:05:15,666 --> 00:05:18,799
so weather decides to keep using deep on
92
00:05:18,900 --> 00:05:21,033
The Lovely Bones in 2,009
93
00:05:21,200 --> 00:05:25,266
and then in a larger capacity on Avatar
94
00:05:26,966 --> 00:05:32,799
now its use on Avatar gets a lot of attention and
95
00:05:34,000 --> 00:05:36,066
people start hearing about you compositing
96
00:05:36,266 --> 00:05:41,233
and this leads to Johanna Sam who
97
00:05:43,700 --> 00:05:46,500
has been aware of Collins work
98
00:05:46,733 --> 00:05:50,966
and he decides to write implementation of
99
00:05:50,966 --> 00:05:53,999
of deep in nuke as nuke 5
100
00:05:55,166 --> 00:05:58,866
and so what Johannes does is
101
00:06:00,133 --> 00:06:04,466
he makes his own nuke plugins
102
00:06:05,133 --> 00:06:06,566
so he has deep nodes
103
00:06:06,900 --> 00:06:09,666
and he publishes videos of him using them
104
00:06:09,666 --> 00:06:12,699
and demonstrating how deep compositing works online
105
00:06:13,300 --> 00:06:14,566
so now everyone
106
00:06:14,866 --> 00:06:16,899
in the industry gets to see how deep compositing works
107
00:06:16,900 --> 00:06:19,500
before they had just heard about it
108
00:06:19,500 --> 00:06:22,666
unless they were witnessing at firsthand at UETA
109
00:06:22,766 --> 00:06:25,066
so now deep compositing becomes
110
00:06:26,533 --> 00:06:30,566
sort of a big deal because now people see the benefits
111
00:06:33,333 --> 00:06:35,433
so what Johannes and Collin
112
00:06:36,666 --> 00:06:38,166
and two other people
113
00:06:38,166 --> 00:06:40,599
Chris Cooper and Daniel Heckenberg
114
00:06:40,600 --> 00:06:44,800
what they do is they decide to write a cigarette paper
115
00:06:44,900 --> 00:06:47,666
that describes deep compositing and
116
00:06:49,666 --> 00:06:50,799
surprisingly
117
00:06:51,166 --> 00:06:55,233
this paper gets rejected by Cigraph
118
00:06:55,366 --> 00:06:56,499
they decide that
119
00:06:56,500 --> 00:06:57,733
it's not gonna be part of the conference
120
00:06:57,733 --> 00:07:00,666
because they say that it's too similar to
121
00:07:00,900 --> 00:07:02,000
Pixar's technology
122
00:07:02,000 --> 00:07:02,800
and it's not
123
00:07:02,966 --> 00:07:04,966
it's not an important development
124
00:07:05,366 --> 00:07:07,733
now the irony here of course
125
00:07:07,733 --> 00:07:09,199
is that deep compositing
126
00:07:09,300 --> 00:07:12,033
ends up becoming a very popular technique
127
00:07:13,666 --> 00:07:14,799
but this paper
128
00:07:16,000 --> 00:07:18,066
is the first time that
129
00:07:18,133 --> 00:07:20,599
people really get to see deep compositing described
130
00:07:20,600 --> 00:07:22,100
from a technical standpoint
131
00:07:23,966 --> 00:07:26,066
so what now
132
00:07:26,533 --> 00:07:28,133
continues to use deep compositing
133
00:07:28,133 --> 00:07:31,599
but they now switch over to nuke in their pipeline
134
00:07:32,066 --> 00:07:35,733
and they use it to a large
135
00:07:35,733 --> 00:07:38,599
large degree on rise of the Final Apes in 2,011
136
00:07:38,733 --> 00:07:42,033
and they get a lot of attention by making a
137
00:07:42,800 --> 00:07:44,966
deep compositing demo video
138
00:07:44,966 --> 00:07:47,966
which some of you may have seen about their work
139
00:07:48,200 --> 00:07:50,433
on rise of the planet of the apes using deep
140
00:07:50,666 --> 00:07:51,666
now at this time
141
00:07:51,666 --> 00:07:55,566
Weta is still using their custom deep pipeline
142
00:07:55,566 --> 00:07:59,899
they're using a format for deep called Open EXR
143
00:08:00,400 --> 00:08:01,800
Deep Z or ODZ
144
00:08:02,066 --> 00:08:05,499
probably ODZ is how they pronounce it
145
00:08:05,766 --> 00:08:10,499
and they use this because they didn't wanna be tied to
146
00:08:10,533 --> 00:08:12,133
the render man native format
147
00:08:12,133 --> 00:08:14,166
they wanted a format that they could use
148
00:08:14,166 --> 00:08:15,699
for other rendering engines as well
149
00:08:15,700 --> 00:08:17,700
and a way to store it um
150
00:08:17,700 --> 00:08:19,366
in an open source format
151
00:08:21,733 --> 00:08:23,499
so at this point
152
00:08:24,300 --> 00:08:30,500
Pixar decides to add deep image output as a feature
153
00:08:30,500 --> 00:08:31,800
natively in Render Man
154
00:08:31,800 --> 00:08:33,500
before it was still a hack
155
00:08:33,533 --> 00:08:37,566
it was still just taking d text deep shadow maps
156
00:08:37,566 --> 00:08:40,299
and using them from the camera's perspective for images
157
00:08:41,100 --> 00:08:43,066
now they add official sport for it
158
00:08:43,066 --> 00:08:44,899
in Render Man Pro Server 16
159
00:08:45,200 --> 00:08:47,566
of course Render Man still uses the DTX format
160
00:08:49,533 --> 00:08:50,866
so now at this point
161
00:08:51,933 --> 00:08:53,799
after Pixar adds official support
162
00:08:53,866 --> 00:08:54,666
the foundry
163
00:08:54,666 --> 00:08:59,166
in Nuke 6.3 decides to add native deep support to nuke
164
00:09:00,266 --> 00:09:04,399
so they make their own deep Nodes
165
00:09:05,333 --> 00:09:07,799
and the first iteration of deep Nodes
166
00:09:08,100 --> 00:09:11,300
only really supported Render Man's Dtex files
167
00:09:13,066 --> 00:09:16,199
they didn't support what they do now
168
00:09:18,066 --> 00:09:21,066
and a lot of studios that used
169
00:09:22,600 --> 00:09:25,166
nuke decided to also started
170
00:09:25,166 --> 00:09:27,066
start to use deep compositing
171
00:09:27,133 --> 00:09:29,166
so now deep compositing is really
172
00:09:29,733 --> 00:09:31,399
exploding in the industry
173
00:09:33,966 --> 00:09:35,233
and critically
174
00:09:36,766 --> 00:09:38,899
the open E XR file format
175
00:09:40,200 --> 00:09:41,500
gets to version 2
176
00:09:41,766 --> 00:09:44,166
and a major part of their update was adding support
177
00:09:44,166 --> 00:09:45,366
for deep image data
178
00:09:46,800 --> 00:09:48,933
so now it's not just Detex files
179
00:09:48,933 --> 00:09:51,333
or other proprietary deep formats
180
00:09:51,333 --> 00:09:54,666
now there's one single format that pipelines can use
181
00:09:55,666 --> 00:09:57,399
to support their deep image data
182
00:09:59,800 --> 00:10:01,200
and in Nuke 8
183
00:10:01,200 --> 00:10:04,200
the foundry had support for Open ex R 2.0
184
00:10:05,000 --> 00:10:06,766
they also had support for
185
00:10:07,000 --> 00:10:09,166
rendering deep image data with the scanline render
186
00:10:09,166 --> 00:10:11,199
and outputting deep data with deep right
187
00:10:11,200 --> 00:10:12,433
prior to this
188
00:10:13,266 --> 00:10:15,466
you could only import data into nuke
189
00:10:15,466 --> 00:10:18,066
you could not output deep data from nuke
190
00:10:18,066 --> 00:10:20,733
so it was really just a one way in
191
00:10:20,733 --> 00:10:21,666
but not out
192
00:10:23,333 --> 00:10:26,366
now go back to the irony of the cigarette paper
193
00:10:26,366 --> 00:10:29,899
being rejected in 2,014
194
00:10:30,966 --> 00:10:32,399
the developers of Deep Compositing
195
00:10:32,400 --> 00:10:34,466
actually won a Technical Achievement award
196
00:10:34,466 --> 00:10:35,766
or a Psychtech Oscar
197
00:10:36,333 --> 00:10:40,699
and this was a huge recognition for them
198
00:10:40,700 --> 00:10:43,200
and it really underscores the contribution
199
00:10:43,200 --> 00:10:47,866
that the industry feels Deep Compositing has had
200
00:10:49,966 --> 00:10:53,233
now today deep compositing is
201
00:10:54,100 --> 00:10:58,933
fairly common in larger studios and high end pipelines
202
00:10:58,933 --> 00:11:01,266
and is supported by most major rendering engines
203
00:11:01,266 --> 00:11:03,433
things like Arnold Houdini's Mantra
204
00:11:03,700 --> 00:11:04,766
Maxwell render
205
00:11:05,000 --> 00:11:07,133
the new red shift pixels
206
00:11:07,133 --> 00:11:08,299
render man of course
207
00:11:08,400 --> 00:11:09,900
V ray through the light
208
00:11:10,766 --> 00:11:11,999
uh there's a few rendering engines
209
00:11:12,000 --> 00:11:13,300
that is not supported by
210
00:11:13,566 --> 00:11:16,399
uh mainly uh
211
00:11:16,400 --> 00:11:18,300
the largest one being Mental Ray
212
00:11:19,766 --> 00:11:21,766
uh and from a compositing standpoint
213
00:11:21,766 --> 00:11:23,799
there are really only two applications
214
00:11:24,166 --> 00:11:25,699
that support deep compositing
215
00:11:25,700 --> 00:11:26,366
and that's of course
216
00:11:26,366 --> 00:11:28,233
nuke and then fusion
217
00:11:28,300 --> 00:11:32,366
um no other 2D applications support deep compositing
218
00:11:32,366 --> 00:11:34,166
so it's still fairly
219
00:11:35,666 --> 00:11:38,999
dedicated to high end pipelines but
220
00:11:41,400 --> 00:11:42,400
maybe in the future
221
00:11:42,400 --> 00:11:45,300
things will spread out and you'll have more options
222
00:11:45,300 --> 00:11:48,933
but we're gonna focus on a nuke in this course
223
00:11:48,933 --> 00:11:51,866
and hopefully this history lesson
224
00:11:51,900 --> 00:11:53,866
is gonna help put things into context
225
00:11:53,866 --> 00:11:56,233
about why certain things are the way they are
226
00:11:56,300 --> 00:11:57,933
why formats are the way they are
227
00:11:57,933 --> 00:12:00,033
why work why workflows are the way they are
228
00:12:00,166 --> 00:12:03,899
and hopefully in your experiences in
229
00:12:03,933 --> 00:12:04,933
in different studios
230
00:12:04,933 --> 00:12:06,699
you might understand sort of
231
00:12:06,900 --> 00:12:08,333
how they got to the pipeline
232
00:12:08,333 --> 00:12:09,399
they got to
16750
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.