Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:06,500 --> 00:00:09,300
hello and welcome back to nuke 2 3 1
2
00:00:09,300 --> 00:00:11,600
nuke composity best practices
3
00:00:11,600 --> 00:00:12,966
this is Victor Perez
4
00:00:13,000 --> 00:00:13,900
and in this class
5
00:00:13,900 --> 00:00:15,666
which is class No. 9
6
00:00:15,666 --> 00:00:17,366
we are going to start
7
00:00:17,566 --> 00:00:21,199
talking about optimization of the 3D renders
8
00:00:21,200 --> 00:00:25,166
so we're going to talk about this can line render note
9
00:00:25,166 --> 00:00:31,199
okay so let's start by one of the most obvious things
10
00:00:31,300 --> 00:00:34,966
which is the format of the skyline render
11
00:00:34,966 --> 00:00:38,566
okay you know that the skyline render has three inputs
12
00:00:38,566 --> 00:00:40,966
one for the object or the scene
13
00:00:41,333 --> 00:00:44,266
the camera and then there is the mysterious input
14
00:00:44,266 --> 00:00:46,099
of the background well
15
00:00:46,266 --> 00:00:48,066
make no mistake here
16
00:00:48,066 --> 00:00:50,799
because the background is not defined
17
00:00:50,800 --> 00:00:54,366
to put the actual background of the composition
18
00:00:54,366 --> 00:00:55,599
even if you can
19
00:00:55,666 --> 00:00:57,499
then if you do that
20
00:00:57,666 --> 00:00:59,566
you cannot separate
21
00:00:59,800 --> 00:01:02,766
the object you are rendering from the background
22
00:01:02,766 --> 00:01:04,699
so we used to put
23
00:01:04,966 --> 00:01:06,399
ah reformat
24
00:01:06,400 --> 00:01:09,733
so the only information this background input
25
00:01:09,733 --> 00:01:12,899
is going to grab from this reformat node
26
00:01:13,133 --> 00:01:15,866
is the actual format
27
00:01:16,133 --> 00:01:20,166
just to output my image in this format
28
00:01:20,500 --> 00:01:24,100
what if you use another note well
29
00:01:24,333 --> 00:01:26,599
there is many people that
30
00:01:26,600 --> 00:01:28,633
instead of using the reformat
31
00:01:28,700 --> 00:01:33,633
they prefer to use the constant
32
00:01:33,900 --> 00:01:37,066
but this is wrong why
33
00:01:37,600 --> 00:01:40,000
look at the boundary box
34
00:01:41,100 --> 00:01:45,366
in theory this reformat
35
00:01:45,966 --> 00:01:50,433
and this constant have the same information
36
00:01:50,466 --> 00:01:51,666
but it's not true
37
00:01:51,666 --> 00:01:53,733
because in the reformat
38
00:01:53,733 --> 00:01:58,599
you get the format of the image with nothing inside
39
00:01:58,766 --> 00:02:00,199
and for nothing
40
00:02:00,200 --> 00:02:03,200
I mean there is no information
41
00:02:03,200 --> 00:02:04,700
as you can see here
42
00:02:04,700 --> 00:02:08,500
I'm just grabbing the information with the color picker
43
00:02:08,500 --> 00:02:09,500
as you can see
44
00:02:09,666 --> 00:02:13,066
but here there is nothing
45
00:02:13,066 --> 00:02:16,366
while in the constant
46
00:02:17,066 --> 00:02:17,933
when you pick
47
00:02:17,933 --> 00:02:20,666
in here you have information
48
00:02:20,800 --> 00:02:23,466
but that information is zero
49
00:02:24,133 --> 00:02:26,033
but it's not the same
50
00:02:26,400 --> 00:02:28,000
have no information
51
00:02:28,000 --> 00:02:29,733
to have information
52
00:02:29,733 --> 00:02:32,299
equals to 0 in every pixel
53
00:02:32,300 --> 00:02:33,500
because in here
54
00:02:33,500 --> 00:02:34,700
your boundary box
55
00:02:34,766 --> 00:02:38,033
is going to process every single pixel
56
00:02:38,166 --> 00:02:40,433
and to be honest with you
57
00:02:40,500 --> 00:02:41,833
nuke doesn't care
58
00:02:41,900 --> 00:02:44,933
if that information is equals to 0
59
00:02:44,933 --> 00:02:47,133
or equals to 2000
60
00:02:47,133 --> 00:02:47,799
is the same
61
00:02:47,800 --> 00:02:50,300
that this information is going to be calculated
62
00:02:50,600 --> 00:02:54,200
so when you are applying the format
63
00:02:54,200 --> 00:02:55,833
to the scanline render
64
00:02:55,900 --> 00:02:58,566
never use a constant
65
00:02:58,866 --> 00:03:01,199
use a reformat instead
66
00:03:01,200 --> 00:03:04,500
okay so that is going to limit the process
67
00:03:04,500 --> 00:03:06,133
of your scan line render
68
00:03:06,133 --> 00:03:06,899
which is an old
69
00:03:06,900 --> 00:03:10,000
that sadly is quite slow to render
70
00:03:10,333 --> 00:03:12,466
and is going to limit that
71
00:03:12,900 --> 00:03:15,466
to this boundary box
72
00:03:15,700 --> 00:03:18,566
which is defining the geometry
73
00:03:18,566 --> 00:03:21,399
okay so remember that the boundary box
74
00:03:21,400 --> 00:03:22,466
in a projection
75
00:03:22,466 --> 00:03:23,499
for instance
76
00:03:23,500 --> 00:03:25,766
is not matching the alpha
77
00:03:25,766 --> 00:03:27,299
or the boundary box
78
00:03:27,300 --> 00:03:29,400
of the image you are projecting
79
00:03:29,400 --> 00:03:31,666
the boundary marks from the scanline render
80
00:03:31,666 --> 00:03:33,799
is always going to match
81
00:03:34,133 --> 00:03:35,099
the geometry
82
00:03:35,100 --> 00:03:37,100
that is going to be rendered out
83
00:03:37,133 --> 00:03:40,066
okay so if you see here
84
00:03:40,800 --> 00:03:42,166
the same image
85
00:03:42,200 --> 00:03:44,033
has no boundary box
86
00:03:44,566 --> 00:03:46,966
is all to the format
87
00:03:47,133 --> 00:03:49,933
okay so another thing is
88
00:03:49,933 --> 00:03:53,199
why not to put anything in there
89
00:03:53,200 --> 00:03:55,700
so leave the input open
90
00:03:55,733 --> 00:03:57,499
well if you do that
91
00:03:57,500 --> 00:03:59,066
and then by any chance
92
00:03:59,066 --> 00:04:01,466
you copy and paste this note
93
00:04:01,466 --> 00:04:02,733
to another project
94
00:04:02,733 --> 00:04:04,966
that is using another project
95
00:04:04,966 --> 00:04:07,966
settings for the full format size
96
00:04:08,366 --> 00:04:10,999
then your skyline render is going to change
97
00:04:11,000 --> 00:04:13,933
and everything else is gonna change
98
00:04:13,933 --> 00:04:15,866
so it's not a good idea
99
00:04:15,866 --> 00:04:19,266
to leave the background input open
100
00:04:19,266 --> 00:04:21,466
you need to put something
101
00:04:21,466 --> 00:04:25,166
and when you put the reformat like here
102
00:04:25,400 --> 00:04:27,166
don't leave the reformat
103
00:04:27,333 --> 00:04:28,966
at the root format
104
00:04:29,000 --> 00:04:31,933
instead use a format in here
105
00:04:31,933 --> 00:04:33,466
so by default
106
00:04:33,566 --> 00:04:34,533
the root format
107
00:04:34,533 --> 00:04:35,866
is going to take
108
00:04:36,100 --> 00:04:37,733
the format of your Priyat
109
00:04:37,733 --> 00:04:40,166
so if your Priyat is set
110
00:04:40,166 --> 00:04:41,299
like you here
111
00:04:41,500 --> 00:04:44,266
at 2K Super 35
112
00:04:44,266 --> 00:04:46,099
for aperture yes
113
00:04:46,366 --> 00:04:47,533
put it in here
114
00:04:47,533 --> 00:04:48,799
don't leave it
115
00:04:48,800 --> 00:04:49,700
to the root
116
00:04:49,700 --> 00:04:51,900
because if you put a reformat
117
00:04:51,900 --> 00:04:52,800
and then you leave
118
00:04:52,800 --> 00:04:53,666
to the root
119
00:04:53,666 --> 00:04:56,366
is exactly the same as putting
120
00:04:56,366 --> 00:04:57,866
nothing okay
121
00:04:57,866 --> 00:05:00,299
so you have to specify
122
00:05:00,333 --> 00:05:02,599
the format in here okay
123
00:05:03,333 --> 00:05:06,033
so that is going to create
124
00:05:06,100 --> 00:05:07,866
a right format
125
00:05:07,866 --> 00:05:09,099
that is going to
126
00:05:09,300 --> 00:05:11,266
be perfectly fit
127
00:05:11,400 --> 00:05:12,800
with the background plate
128
00:05:12,800 --> 00:05:15,433
and then you just need to put a merge
129
00:05:15,533 --> 00:05:17,466
and put this over
130
00:05:17,466 --> 00:05:21,199
or any other operation of your choice okay
131
00:05:22,800 --> 00:05:25,700
now another thing that is very interesting
132
00:05:25,700 --> 00:05:28,533
when we are working with the skyline render
133
00:05:28,533 --> 00:05:34,899
is sometimes we have an image like this okay um
134
00:05:34,900 --> 00:05:40,166
maybe you want to apply a certain process to this image
135
00:05:40,166 --> 00:05:42,466
okay that process can be
136
00:05:42,466 --> 00:05:43,366
for instance
137
00:05:43,366 --> 00:05:44,799
a color correction
138
00:05:44,800 --> 00:05:46,533
or a convolution
139
00:05:46,533 --> 00:05:47,566
for creating
140
00:05:47,566 --> 00:05:48,333
for instance
141
00:05:48,333 --> 00:05:52,299
a glow or any other process you want to apply
142
00:05:52,300 --> 00:05:54,700
so how to do that
143
00:05:54,700 --> 00:05:57,266
well depending on the operation
144
00:05:57,333 --> 00:06:00,733
you want to use one method or another
145
00:06:00,733 --> 00:06:04,566
okay so imagine that I want to create
146
00:06:04,866 --> 00:06:07,366
some kind of color correction
147
00:06:07,366 --> 00:06:10,099
okay so if I have this
148
00:06:10,666 --> 00:06:11,899
and this is my alpha
149
00:06:11,900 --> 00:06:14,500
there is no way you can separate that
150
00:06:14,966 --> 00:06:16,433
but remember
151
00:06:16,666 --> 00:06:20,266
the scanning render is going to render every channel
152
00:06:20,666 --> 00:06:26,033
you are putting inside the UVs of your geometry
153
00:06:26,166 --> 00:06:28,099
saying with the predictions okay
154
00:06:28,166 --> 00:06:35,399
so if you having here a prediction of gas r g b a
155
00:06:35,700 --> 00:06:40,033
the output result is going to be RGBA
156
00:06:40,700 --> 00:06:43,166
but if I want to isolate
157
00:06:43,466 --> 00:06:44,599
for instance
158
00:06:44,733 --> 00:06:47,999
this first sphere okay
159
00:06:48,533 --> 00:06:52,166
that is corresponding to this okay
160
00:06:53,300 --> 00:06:59,700
what I want is just to create kind of ID pass remember
161
00:06:59,700 --> 00:07:03,633
the ID pass is the pass that is going to give me a mat
162
00:07:03,800 --> 00:07:06,600
for a certain element or geometry
163
00:07:06,600 --> 00:07:08,933
okay so how to create that
164
00:07:08,933 --> 00:07:10,499
well it's going to be really
165
00:07:10,500 --> 00:07:11,366
really easy
166
00:07:11,666 --> 00:07:13,766
I'm going to call a shuffle note
167
00:07:14,200 --> 00:07:17,266
and then I'm going to place that shuffle note in here
168
00:07:17,466 --> 00:07:18,999
and in this case in
169
00:07:19,000 --> 00:07:23,800
instead of taking any pass from my original image
170
00:07:23,800 --> 00:07:26,033
and then placing into the RGB
171
00:07:26,266 --> 00:07:27,699
what I'm going to do is
172
00:07:27,700 --> 00:07:32,066
I'm going to take one output layer of my choice
173
00:07:32,066 --> 00:07:32,699
that can be
174
00:07:32,700 --> 00:07:33,366
for instance
175
00:07:33,366 --> 00:07:34,533
the mask or
176
00:07:34,533 --> 00:07:35,533
if you prefer
177
00:07:35,533 --> 00:07:37,299
you can create it yourself
178
00:07:37,300 --> 00:07:38,533
so in this case
179
00:07:38,533 --> 00:07:39,933
I'm going to put the mask
180
00:07:39,933 --> 00:07:42,433
because I just want one single channel
181
00:07:42,500 --> 00:07:48,100
and now I'm going to feel that mask with full 1 value
182
00:07:48,100 --> 00:07:49,500
so if you look at this
183
00:07:49,500 --> 00:07:50,866
this is RGB
184
00:07:51,000 --> 00:07:52,466
now this is the a
185
00:07:52,466 --> 00:07:53,666
okay and now
186
00:07:53,666 --> 00:07:55,033
if I go in here
187
00:07:55,700 --> 00:07:57,466
the mask here you are
188
00:07:57,466 --> 00:07:59,666
this is the information of my mask
189
00:07:59,666 --> 00:08:00,666
that is display
190
00:08:00,666 --> 00:08:01,966
by the way in red
191
00:08:02,100 --> 00:08:05,166
just because it's the only channel available
192
00:08:05,166 --> 00:08:06,066
but it's not that
193
00:08:06,066 --> 00:08:07,566
my mask is red
194
00:08:07,566 --> 00:08:11,099
is I'm just caring about one single channel
195
00:08:11,100 --> 00:08:13,900
and the first channel available in nuke
196
00:08:14,066 --> 00:08:14,799
he's the red one
197
00:08:14,800 --> 00:08:19,100
so that's why the aparency of my mask is red
198
00:08:19,300 --> 00:08:22,866
so let's go back to the RGBA set
199
00:08:23,266 --> 00:08:24,699
okay and now
200
00:08:24,700 --> 00:08:28,366
if I see my output
201
00:08:28,366 --> 00:08:29,966
you will see our GBA
202
00:08:30,733 --> 00:08:34,099
okay but attention
203
00:08:34,100 --> 00:08:35,600
what is happening in here
204
00:08:35,866 --> 00:08:37,966
now I have an extra channel
205
00:08:37,966 --> 00:08:41,099
okay that extra channel is the mask
206
00:08:41,500 --> 00:08:45,833
but he's being applied only to this geometry
207
00:08:45,966 --> 00:08:48,266
so if I go here
208
00:08:48,266 --> 00:08:49,266
to the mask
209
00:08:49,800 --> 00:08:50,833
here you are
210
00:08:51,000 --> 00:08:53,500
so now that mask is isolated
211
00:08:53,500 --> 00:08:56,266
so if I want to apply a certain effect
212
00:08:57,700 --> 00:08:59,700
I just need to go here
213
00:09:00,466 --> 00:09:03,599
great now I'm going to use the mask
214
00:09:03,866 --> 00:09:08,433
okay so now I'm going to create that certain effect
215
00:09:08,866 --> 00:09:11,633
for that piece of geometry okay
216
00:09:12,466 --> 00:09:17,866
but that is actually not perfectly accurate
217
00:09:18,300 --> 00:09:21,600
when I want to create a color correction
218
00:09:22,800 --> 00:09:24,833
can you see what is happening in here
219
00:09:25,900 --> 00:09:31,766
the edge is creating that glowy effect
220
00:09:31,966 --> 00:09:35,266
and that's because I'm using a mask
221
00:09:35,266 --> 00:09:37,733
that is matching the alpha
222
00:09:37,733 --> 00:09:42,199
so it's exactly the same effect of applying
223
00:09:42,200 --> 00:09:44,966
instead of a mask for the ID
224
00:09:44,966 --> 00:09:47,299
I'm going to apply the mask for the alpha
225
00:09:47,300 --> 00:09:50,700
so now the same effect is happening for everything
226
00:09:50,733 --> 00:09:52,266
and as you know
227
00:09:52,300 --> 00:09:55,500
you never use the alpha as a mask
228
00:09:55,666 --> 00:10:00,299
the alpha must be used as a pre multiplication process
229
00:10:00,300 --> 00:10:02,200
okay so you unpro multiply
230
00:10:02,200 --> 00:10:04,533
you then apply the process
231
00:10:04,533 --> 00:10:06,699
and then you pre multiply
232
00:10:06,966 --> 00:10:09,833
this is something that you cannot accomplish
233
00:10:09,966 --> 00:10:10,966
with this process
234
00:10:10,966 --> 00:10:14,066
because if you use the mask for um
235
00:10:14,066 --> 00:10:15,266
primitiplication
236
00:10:15,266 --> 00:10:19,366
something like here the mask
237
00:10:19,700 --> 00:10:22,766
then everything is going to disappear
238
00:10:22,900 --> 00:10:25,500
because he's applying a non primal duplication
239
00:10:25,500 --> 00:10:28,600
and then the primal duplication by this shape
240
00:10:28,733 --> 00:10:31,099
which is just creating that effect
241
00:10:31,133 --> 00:10:33,499
and even if you force that
242
00:10:33,800 --> 00:10:34,966
and you for instance
243
00:10:34,966 --> 00:10:38,566
instead of doing that automatically inside the node
244
00:10:38,566 --> 00:10:40,633
you prefer to do it manually with nodes
245
00:10:40,700 --> 00:10:42,766
so I'm going to call it here
246
00:10:43,400 --> 00:10:44,500
the amprimoult
247
00:10:44,766 --> 00:10:48,999
so I'm going to amprimoult by the mask
248
00:10:49,466 --> 00:10:53,666
and then to premoult after by the
249
00:10:56,000 --> 00:10:59,366
alpha okay so in here
250
00:11:00,300 --> 00:11:01,233
you get that
251
00:11:01,566 --> 00:11:02,566
and in here
252
00:11:02,933 --> 00:11:04,466
everything looks fine
253
00:11:04,466 --> 00:11:06,499
except those edges
254
00:11:06,500 --> 00:11:07,300
and those edges
255
00:11:07,300 --> 00:11:12,166
so there is no way you can change the color
256
00:11:12,166 --> 00:11:15,199
with that extreme color correction
257
00:11:15,700 --> 00:11:17,400
by using a mask
258
00:11:17,533 --> 00:11:21,099
that mask can be used very nicely for a glow
259
00:11:21,100 --> 00:11:23,200
or for delimitating
260
00:11:23,200 --> 00:11:27,100
any effect of convolutions that it is okay
261
00:11:27,300 --> 00:11:29,300
or to mask other
262
00:11:29,666 --> 00:11:33,799
effects that you can apply to the other spheres
263
00:11:33,800 --> 00:11:37,500
like set blur or something like that
264
00:11:37,700 --> 00:11:41,333
actually this method of isolation
265
00:11:41,333 --> 00:11:44,266
is not great for color correction
266
00:11:44,266 --> 00:11:46,899
okay so how do you do that
267
00:11:46,900 --> 00:11:50,500
in case you want to isolate an element well
268
00:11:50,866 --> 00:11:54,899
that is nothing better than getting your sphere
269
00:11:55,000 --> 00:11:57,600
separate in a different render
270
00:11:57,966 --> 00:11:58,866
but in this case
271
00:11:58,866 --> 00:12:00,099
is going to be really easy
272
00:12:00,100 --> 00:12:03,500
if you apply this fear to be the first fear
273
00:12:03,500 --> 00:12:07,800
so everything is going to be under this fear
274
00:12:07,800 --> 00:12:09,800
in a layer combination okay
275
00:12:09,800 --> 00:12:13,900
but what if this fear I want to change is the green one
276
00:12:14,100 --> 00:12:17,100
well things can get complicated
277
00:12:17,133 --> 00:12:20,566
okay so I'm going to show you how to integrate
278
00:12:20,566 --> 00:12:24,466
deep image compositing to apply that certain effect
279
00:12:24,466 --> 00:12:27,199
so in here I have
280
00:12:27,733 --> 00:12:32,633
the render of my red and the blue spheres
281
00:12:32,966 --> 00:12:37,833
and here I have the render for the green sphere okay
282
00:12:38,133 --> 00:12:40,499
you know that automatically
283
00:12:40,500 --> 00:12:42,200
even if you don't want to
284
00:12:42,800 --> 00:12:48,266
this calendar is releasing the deep information okay
285
00:12:48,266 --> 00:12:54,366
so if you use something as simple as the deep merch
286
00:12:54,366 --> 00:12:55,399
here you are
287
00:12:56,766 --> 00:12:59,866
you are going to get this merch
288
00:13:00,666 --> 00:13:03,599
and you are going to mix this and this
289
00:13:03,666 --> 00:13:04,599
and together
290
00:13:04,666 --> 00:13:05,766
here you are
291
00:13:05,766 --> 00:13:10,399
okay so he's going to respect what is in front
292
00:13:10,466 --> 00:13:15,599
of my green sphere and what is beyond my green sphere
293
00:13:15,866 --> 00:13:17,133
that's very nice
294
00:13:17,133 --> 00:13:18,199
the problem is
295
00:13:18,533 --> 00:13:23,266
as soon as you put a node after this can line render
296
00:13:23,866 --> 00:13:25,099
the information
297
00:13:25,166 --> 00:13:29,033
of the deep image compositing is going to get lost
298
00:13:29,366 --> 00:13:34,233
so how can you manipulate maybe your green sphere
299
00:13:34,800 --> 00:13:39,500
and still maintain this deep image compositing setup
300
00:13:39,533 --> 00:13:41,699
well for that
301
00:13:41,966 --> 00:13:45,766
we need to get a little work around
302
00:13:45,766 --> 00:13:49,799
using another note of deep image compositing
303
00:13:49,866 --> 00:13:51,966
which is in this case
304
00:13:53,333 --> 00:13:55,699
the deep holdout
305
00:13:56,133 --> 00:14:01,299
so the deep holdout is going to create the hole in here
306
00:14:01,666 --> 00:14:05,699
just to place this green sphere
307
00:14:05,933 --> 00:14:11,299
using an operation that is not the over
308
00:14:11,366 --> 00:14:14,666
okay every time we are using a holdout
309
00:14:14,933 --> 00:14:17,033
we are going to use the
310
00:14:17,266 --> 00:14:18,699
this joint over
311
00:14:18,700 --> 00:14:20,200
I'm going to show you now
312
00:14:20,400 --> 00:14:24,400
but first let's apply the deep holdout
313
00:14:24,700 --> 00:14:29,200
so in here you are going to apply the image
314
00:14:29,200 --> 00:14:34,333
that you want to crop from this image
315
00:14:34,333 --> 00:14:38,166
okay so that is the holdout input
316
00:14:38,333 --> 00:14:40,933
and then because this is my main element
317
00:14:40,933 --> 00:14:44,499
that is going to be feed into the main input
318
00:14:44,733 --> 00:14:47,533
remember that the holdout is a note
319
00:14:47,533 --> 00:14:50,466
that is a processing deep image
320
00:14:50,766 --> 00:14:52,199
but the result
321
00:14:52,200 --> 00:14:56,933
the output is going to be a regular to the image okay
322
00:14:56,933 --> 00:14:59,666
that's why you are getting in this note
323
00:15:00,200 --> 00:15:02,300
the rectangles of the color
324
00:15:02,300 --> 00:15:06,700
so every note that has the rectangles of the colors
325
00:15:06,766 --> 00:15:08,599
corresponding the channels
326
00:15:08,800 --> 00:15:11,266
that is a to the process
327
00:15:11,266 --> 00:15:12,699
okay so in here
328
00:15:12,700 --> 00:15:16,166
if I apply this to this
329
00:15:16,566 --> 00:15:19,399
the result is going to be that
330
00:15:19,566 --> 00:15:23,066
so is the portion of the sphere
331
00:15:23,366 --> 00:15:27,099
that is reveal by the rest of the scene
332
00:15:27,100 --> 00:15:30,266
so the red sphere in front of the green
333
00:15:30,300 --> 00:15:33,700
is now cropping those pixels okay
334
00:15:34,200 --> 00:15:36,333
um before to apply a glow
335
00:15:36,333 --> 00:15:38,399
because this is very simple
336
00:15:38,400 --> 00:15:39,266
you apply a glow
337
00:15:39,266 --> 00:15:42,733
and then you put in everything together
338
00:15:42,733 --> 00:15:46,166
I want you to think about something
339
00:15:46,466 --> 00:15:48,666
remember that we are applying
340
00:15:50,500 --> 00:15:54,766
a mask to crop those pixels
341
00:15:54,800 --> 00:15:59,500
okay so if we put anover
342
00:16:00,466 --> 00:16:05,899
the effect that we are using is we have that
343
00:16:07,133 --> 00:16:08,166
and this
344
00:16:10,800 --> 00:16:15,600
of course in this schematic drawing is very clear
345
00:16:16,266 --> 00:16:20,433
in reality this is happening exactly the same
346
00:16:20,533 --> 00:16:25,699
which is this area is not touching
347
00:16:26,200 --> 00:16:29,100
well it's actually very very
348
00:16:29,100 --> 00:16:30,366
very similar in the value
349
00:16:30,366 --> 00:16:31,933
so if you use an over
350
00:16:31,933 --> 00:16:35,099
probably you're not going to notice much
351
00:16:35,333 --> 00:16:37,999
but as soon as you are getting motion blur
352
00:16:38,000 --> 00:16:39,566
or soft edges
353
00:16:39,733 --> 00:16:44,166
this portion of the edge of the alpha
354
00:16:44,400 --> 00:16:47,366
not being perfectly match
355
00:16:47,500 --> 00:16:50,900
with the image above and below
356
00:16:51,133 --> 00:16:54,066
is going to create a dark edge
357
00:16:54,266 --> 00:16:58,899
okay so if you don't want to have that dark edge
358
00:16:58,966 --> 00:17:02,666
is better that you don't use and over
359
00:17:02,733 --> 00:17:05,799
because the over is designed to put two images
360
00:17:05,800 --> 00:17:10,500
that are just a main image with transparency
361
00:17:10,566 --> 00:17:12,933
over a full cover image
362
00:17:12,933 --> 00:17:13,599
but in here
363
00:17:13,600 --> 00:17:17,000
we have already my image pre cut
364
00:17:17,133 --> 00:17:20,766
so if I put this and this
365
00:17:20,766 --> 00:17:22,433
to deactivate the glow
366
00:17:23,300 --> 00:17:24,400
and to show you
367
00:17:25,266 --> 00:17:27,266
have a look at those pixels
368
00:17:27,900 --> 00:17:30,233
this is kind of pinky
369
00:17:30,500 --> 00:17:32,433
and this is kind of green
370
00:17:32,533 --> 00:17:38,066
so in here we have this tiny
371
00:17:38,133 --> 00:17:41,799
tiny variation in the value that is getting dark
372
00:17:42,200 --> 00:17:47,366
which is because in there you have actually dark values
373
00:17:47,466 --> 00:17:48,566
and in here
374
00:17:48,666 --> 00:17:50,166
you have the same dark value
375
00:17:50,166 --> 00:17:53,299
so when you put those pixels together
376
00:17:53,766 --> 00:17:56,633
you don't get the full amount of value
377
00:17:56,966 --> 00:17:59,699
just to represent the perfect blend
378
00:17:59,733 --> 00:18:02,099
from the pink to the green
379
00:18:02,266 --> 00:18:05,099
so when you have this effect
380
00:18:05,500 --> 00:18:08,766
you have to use the this join over
381
00:18:08,866 --> 00:18:11,966
okay the this join over represents
382
00:18:11,966 --> 00:18:16,466
when the alpha is not perfectly feeling all the gaps
383
00:18:16,600 --> 00:18:18,866
so with the this join over
384
00:18:18,866 --> 00:18:23,966
you're going to get this perfect sense of blending
385
00:18:23,966 --> 00:18:26,833
of an image with the holdout
386
00:18:27,000 --> 00:18:30,200
okay so with this drainover
387
00:18:30,200 --> 00:18:32,600
everything is looking great
388
00:18:32,866 --> 00:18:38,466
so I'm going to check if this and this
389
00:18:38,866 --> 00:18:40,833
which is the deep image
390
00:18:41,200 --> 00:18:42,433
with the merch
391
00:18:42,500 --> 00:18:44,466
and these with the this join
392
00:18:45,000 --> 00:18:46,100
are accurate
393
00:18:46,966 --> 00:18:49,899
okay so here you are
394
00:18:50,400 --> 00:18:51,900
perfectly accurate
395
00:18:52,133 --> 00:18:53,566
from one to the other
396
00:18:54,200 --> 00:18:57,266
okay notice that in here
397
00:18:57,266 --> 00:18:58,333
if I change
398
00:18:58,333 --> 00:19:03,299
and I compare the over and the deep match see
399
00:19:05,166 --> 00:19:09,366
this is the edge I was talking about okay
400
00:19:09,366 --> 00:19:15,433
now that we have that merge perfectly mixing my images
401
00:19:15,466 --> 00:19:17,299
now I can apply the process
402
00:19:17,300 --> 00:19:19,466
and if you want to apply the process
403
00:19:19,533 --> 00:19:21,366
you can apply the glow
404
00:19:21,700 --> 00:19:23,400
and everything is going to work fine
405
00:19:23,400 --> 00:19:26,833
even the pixels that are getting greenish
406
00:19:26,900 --> 00:19:28,566
on the pinky area
407
00:19:29,366 --> 00:19:31,566
but not only in here
408
00:19:31,566 --> 00:19:34,799
I can also apply a great
409
00:19:34,800 --> 00:19:35,566
and in this case
410
00:19:35,566 --> 00:19:39,366
because this image is just a pre multiply image
411
00:19:39,733 --> 00:19:41,633
I can apply in here
412
00:19:42,200 --> 00:19:46,033
yes I'm promote remote from the note
413
00:19:46,100 --> 00:19:49,000
and then I can change anything
414
00:19:50,600 --> 00:19:56,033
and is going to look perfect in every single pixel
415
00:19:56,266 --> 00:19:58,133
okay so this way
416
00:19:58,133 --> 00:20:00,266
using the deep image compositing
417
00:20:00,266 --> 00:20:02,299
is going to speed up your overflow
418
00:20:02,300 --> 00:20:03,700
and is going to avoid you
419
00:20:03,700 --> 00:20:07,500
to create three layers of the skyline render
420
00:20:07,600 --> 00:20:10,433
one from the red sphere
421
00:20:10,733 --> 00:20:12,499
one for the green sphere
422
00:20:12,500 --> 00:20:14,833
and one for the blue sphere
423
00:20:14,866 --> 00:20:15,566
and in this case
424
00:20:15,566 --> 00:20:16,366
is really easy
425
00:20:16,366 --> 00:20:19,799
because they are not touching from one to the other
426
00:20:19,800 --> 00:20:22,166
but when you have things that are going
427
00:20:22,200 --> 00:20:23,566
one in front of the other
428
00:20:23,566 --> 00:20:25,766
and then back behind
429
00:20:25,800 --> 00:20:28,700
then everything is going to be a bit messy
430
00:20:28,700 --> 00:20:32,500
so using deep image compositing
431
00:20:32,500 --> 00:20:34,800
for this kind of holding out
432
00:20:34,800 --> 00:20:37,600
is going to help a lot okay
433
00:20:38,100 --> 00:20:40,733
so that is for layering
434
00:20:40,733 --> 00:20:44,466
using advanced techniques of this calendar
435
00:20:44,500 --> 00:20:45,166
another thing
436
00:20:45,166 --> 00:20:48,866
that we have to keep in consideration is that
437
00:20:50,166 --> 00:20:54,566
when you are rendering an image with motion blur
438
00:20:55,000 --> 00:20:57,066
everybody knows that the motion blur
439
00:20:57,066 --> 00:21:00,099
is actually being created by samples
440
00:21:00,100 --> 00:21:02,266
so I'm going to show you
441
00:21:02,600 --> 00:21:05,800
in the skyline render how the samples work
442
00:21:05,900 --> 00:21:08,966
so here we have the multi sample
443
00:21:09,366 --> 00:21:12,499
so the amount of motion blur
444
00:21:12,500 --> 00:21:13,166
and the amount
445
00:21:13,166 --> 00:21:17,133
I mean how long is the tail of your motion blur
446
00:21:17,133 --> 00:21:19,866
is going to depend in terms of quality
447
00:21:20,100 --> 00:21:24,266
of the amount of samples you are going to apply in here
448
00:21:24,266 --> 00:21:29,466
because the samples are just a render of the same image
449
00:21:29,466 --> 00:21:32,133
in different position of times
450
00:21:32,133 --> 00:21:33,399
in between the frame
451
00:21:33,400 --> 00:21:36,833
so is rendering from frame 1 to frame 2
452
00:21:36,933 --> 00:21:39,499
several different positions in time
453
00:21:39,500 --> 00:21:41,166
so let's say
454
00:21:41,200 --> 00:21:46,100
render 1.1 frame 1.2
455
00:21:46,100 --> 00:21:48,900
frames 1.3 and so on
456
00:21:49,500 --> 00:21:50,866
it's something that is going to
457
00:21:50,866 --> 00:21:53,099
blend in one single frame
458
00:21:53,100 --> 00:21:54,566
so you are going to have
459
00:21:54,733 --> 00:21:58,333
the feeling that that image is actually moving
460
00:21:58,333 --> 00:22:00,566
but is related with the amount of samples
461
00:22:00,566 --> 00:22:03,566
and the amount of samples is as plain as
462
00:22:03,900 --> 00:22:07,900
you are rendering this image 100 times
463
00:22:07,900 --> 00:22:08,900
in this case
464
00:22:08,933 --> 00:22:10,366
if you render twice
465
00:22:11,100 --> 00:22:17,100
that is going to take twice the time for rendering
466
00:22:17,100 --> 00:22:18,966
if you put 5
467
00:22:19,000 --> 00:22:21,166
is going to take five times
468
00:22:21,166 --> 00:22:22,099
as you can see
469
00:22:22,100 --> 00:22:24,966
it's really clear that if you put 3
470
00:22:25,100 --> 00:22:30,600
you can see the shape of my image three times okay
471
00:22:31,166 --> 00:22:33,633
and he's blending everything together
472
00:22:34,366 --> 00:22:36,433
but to get a good effect
473
00:22:36,933 --> 00:22:38,566
depending on the amount of mushroom blood
474
00:22:38,566 --> 00:22:41,333
you are going to need several samples
475
00:22:41,333 --> 00:22:44,999
so in here I decide to put 100
476
00:22:45,800 --> 00:22:48,366
some of you can think now okay
477
00:22:48,400 --> 00:22:52,000
if I render something complex 100 times
478
00:22:52,000 --> 00:22:56,100
that is going to take a lot of time for calculation
479
00:22:56,100 --> 00:22:57,233
and that's true
480
00:22:57,366 --> 00:22:59,566
so then in here
481
00:22:59,566 --> 00:23:02,066
if you don't want to calculate the samples
482
00:23:02,333 --> 00:23:04,966
you can calculate what is called
483
00:23:08,000 --> 00:23:10,300
a motion vector pass
484
00:23:10,566 --> 00:23:11,699
and in nuke
485
00:23:11,866 --> 00:23:14,899
you have different kind of motion vectors
486
00:23:14,900 --> 00:23:17,533
depending on the information you want to represent
487
00:23:17,533 --> 00:23:18,999
okay by default
488
00:23:19,000 --> 00:23:21,200
you are going to get distance okay
489
00:23:21,200 --> 00:23:22,500
which is the distant
490
00:23:22,500 --> 00:23:26,800
that every pixel is representing movement
491
00:23:26,933 --> 00:23:27,999
in this case
492
00:23:28,200 --> 00:23:29,566
forward okay
493
00:23:29,566 --> 00:23:31,399
from one pixel to another
494
00:23:31,400 --> 00:23:34,300
so then once you have that pass
495
00:23:34,300 --> 00:23:38,766
which is looking something like this okay
496
00:23:38,866 --> 00:23:41,199
is representing in u v
497
00:23:41,200 --> 00:23:44,433
which is the relative coordinates x y
498
00:23:45,700 --> 00:23:46,966
he's representing
499
00:23:47,066 --> 00:23:49,766
the distance from this pixel to the next one
500
00:23:49,766 --> 00:23:51,333
intense of movement
501
00:23:51,333 --> 00:23:54,533
so using that information
502
00:23:54,533 --> 00:23:57,566
which is baked in the render
503
00:23:57,600 --> 00:23:58,966
pixel by pixel
504
00:23:59,100 --> 00:24:01,966
I can use what is called a vector blur
505
00:24:02,166 --> 00:24:04,166
the vector blur is in here
506
00:24:04,800 --> 00:24:07,300
and it's going to apply that
507
00:24:07,300 --> 00:24:09,400
a movement of every pixel
508
00:24:10,000 --> 00:24:11,400
leaving the trail
509
00:24:11,400 --> 00:24:13,266
from the original position
510
00:24:13,266 --> 00:24:15,666
to the position of destination
511
00:24:15,666 --> 00:24:18,166
in both X and Y
512
00:24:18,600 --> 00:24:20,566
but keep this in mind
513
00:24:20,666 --> 00:24:23,666
is just X and y
514
00:24:23,800 --> 00:24:25,933
there is noted in here
515
00:24:25,933 --> 00:24:31,499
so you cannot create 3D motion blur in here even
516
00:24:31,500 --> 00:24:33,266
because what you are doing here
517
00:24:33,266 --> 00:24:35,366
is just a compolution of pixels
518
00:24:35,366 --> 00:24:37,699
based on a matrics of movement
519
00:24:37,933 --> 00:24:39,199
but is just
520
00:24:39,200 --> 00:24:43,700
that is very different from getting the image render
521
00:24:43,700 --> 00:24:47,133
in several different positions in time
522
00:24:47,133 --> 00:24:48,899
between one frame to another
523
00:24:48,900 --> 00:24:52,400
in here is just plain movement into the
524
00:24:52,500 --> 00:24:57,200
so when you apply those motion vectors to the image
525
00:24:58,000 --> 00:25:00,033
you get something like that
526
00:25:00,333 --> 00:25:07,466
which is more or less similar to the final image okay
527
00:25:08,366 --> 00:25:09,566
as you can see
528
00:25:10,066 --> 00:25:11,699
there are artifacts
529
00:25:11,700 --> 00:25:14,100
and that's normal okay
530
00:25:14,100 --> 00:25:15,166
with the vector blur
531
00:25:15,166 --> 00:25:18,299
I apply a certain channel
532
00:25:19,000 --> 00:25:21,066
then I use my alpha to define
533
00:25:21,066 --> 00:25:25,866
what part of my image needs to be moved okay
534
00:25:25,866 --> 00:25:29,333
and then I was applying the amplitude of my movement
535
00:25:29,333 --> 00:25:32,866
so I don't care about the setup of this vector blur
536
00:25:32,866 --> 00:25:36,433
what I care is about the process itself
537
00:25:36,500 --> 00:25:37,733
and the process
538
00:25:37,733 --> 00:25:38,999
let me tell you that
539
00:25:39,000 --> 00:25:41,966
calculating the motion blur
540
00:25:41,966 --> 00:25:45,599
using the vector blur is not a great
541
00:25:45,933 --> 00:25:49,299
sometimes you can be lucky and don't get
542
00:25:49,800 --> 00:25:51,566
those artifacts
543
00:25:51,966 --> 00:25:56,266
but most of the time is kind of dangerous to apply that
544
00:25:56,266 --> 00:25:59,999
and even if the result is quite similar
545
00:26:00,566 --> 00:26:04,999
the quality of the motion blur from the 3D
546
00:26:05,500 --> 00:26:08,866
and the vector blur is so different
547
00:26:08,866 --> 00:26:12,899
there is no way you can even compare the result in here
548
00:26:12,900 --> 00:26:15,100
is is just a process of today
549
00:26:15,100 --> 00:26:17,400
but your brain is going to be very smart
550
00:26:17,400 --> 00:26:18,533
and is going to understand
551
00:26:18,533 --> 00:26:21,966
when that motion blur is real or not
552
00:26:22,200 --> 00:26:25,266
even if in the skyline render
553
00:26:25,266 --> 00:26:27,599
the motion blur that we are calculating
554
00:26:27,600 --> 00:26:30,533
is just a linear fashion motion blur
555
00:26:30,533 --> 00:26:34,299
so it's a mathematical way of representing
556
00:26:34,366 --> 00:26:36,999
the soup information from the time
557
00:26:37,133 --> 00:26:38,166
while in here
558
00:26:38,166 --> 00:26:39,099
you have just
559
00:26:39,200 --> 00:26:40,300
in this case
560
00:26:40,300 --> 00:26:42,666
24 segments of time
561
00:26:42,666 --> 00:26:44,533
but we are inventing
562
00:26:44,533 --> 00:26:47,799
the information that is in between those frames
563
00:26:48,566 --> 00:26:51,999
the vector blur is not a good idea
564
00:26:52,133 --> 00:26:53,899
just try it
565
00:26:54,066 --> 00:26:56,699
but just a tiny amount
566
00:26:56,700 --> 00:26:58,400
because the vector blur
567
00:26:59,133 --> 00:27:00,799
he's not going to be used
568
00:27:00,800 --> 00:27:03,866
in general I like to use the vector blur
569
00:27:03,866 --> 00:27:06,299
and to apply the motion blur
570
00:27:06,300 --> 00:27:09,766
when I'm applying a patch over a CG
571
00:27:09,866 --> 00:27:11,099
and then I need to
572
00:27:11,100 --> 00:27:14,100
create the movement of that patch in the CG
573
00:27:14,300 --> 00:27:15,200
with the same
574
00:27:15,200 --> 00:27:15,966
more or less
575
00:27:15,966 --> 00:27:18,666
fashion of the movement okay
576
00:27:18,700 --> 00:27:20,000
but by default
577
00:27:20,000 --> 00:27:22,966
I would never recommend you to use
578
00:27:23,166 --> 00:27:27,199
the vector blur for general motion blur okay
579
00:27:27,600 --> 00:27:29,866
still if you want to get better
580
00:27:29,866 --> 00:27:31,933
results with your motion blur
581
00:27:31,933 --> 00:27:33,866
I'm going to give you an advice
582
00:27:34,400 --> 00:27:36,000
which is in here
583
00:27:36,466 --> 00:27:38,666
you have this randomized time
584
00:27:38,666 --> 00:27:40,099
I'm going to reduce
585
00:27:40,966 --> 00:27:43,199
my samples by 20
586
00:27:44,200 --> 00:27:46,966
so you are going to see clearly
587
00:27:47,300 --> 00:27:48,233
the pattern
588
00:27:48,300 --> 00:27:50,400
of the linear representation
589
00:27:50,400 --> 00:27:51,266
I mean linear
590
00:27:51,266 --> 00:27:53,799
because the distance between all samples
591
00:27:53,800 --> 00:27:56,133
is arapematically the same
592
00:27:56,133 --> 00:27:59,333
okay so in here
593
00:27:59,333 --> 00:28:01,166
everything is looking like
594
00:28:01,166 --> 00:28:02,999
mathematically perfect
595
00:28:03,400 --> 00:28:05,100
but you can apply
596
00:28:05,400 --> 00:28:08,333
what is called a randomization of the time
597
00:28:08,333 --> 00:28:10,033
so is going to take
598
00:28:10,266 --> 00:28:11,666
different distance
599
00:28:11,666 --> 00:28:13,299
for every sample
600
00:28:13,333 --> 00:28:15,599
so that is going to hide
601
00:28:15,666 --> 00:28:17,366
the effect of Muay
602
00:28:17,366 --> 00:28:17,999
for instance
603
00:28:18,000 --> 00:28:19,466
which is those lines
604
00:28:19,466 --> 00:28:20,666
doing weird things
605
00:28:20,666 --> 00:28:22,566
when they duplicate
606
00:28:22,566 --> 00:28:23,633
like in here
607
00:28:24,066 --> 00:28:26,766
and that is going to hide many problems
608
00:28:26,766 --> 00:28:29,199
of course you are not going to resolve
609
00:28:29,200 --> 00:28:31,366
hold the problems of the samples
610
00:28:31,500 --> 00:28:33,733
but my aiming here
611
00:28:33,733 --> 00:28:36,666
is how to provide you the information
612
00:28:36,900 --> 00:28:38,600
to render your image
613
00:28:38,600 --> 00:28:40,433
to look fine
614
00:28:40,466 --> 00:28:45,433
without having a crazy amount of samples
615
00:28:45,500 --> 00:28:47,900
so if we randomize the time
616
00:28:50,800 --> 00:28:55,000
we can create that irregularity
617
00:28:55,133 --> 00:28:56,566
okay so in here
618
00:28:56,566 --> 00:28:58,466
you can see very clear in those lines
619
00:28:58,466 --> 00:29:01,299
that the space between the lines
620
00:29:02,133 --> 00:29:05,599
is not the same as when you were in here
621
00:29:05,600 --> 00:29:07,200
see that everything
622
00:29:07,366 --> 00:29:11,133
all the distances are quite the same
623
00:29:11,133 --> 00:29:13,599
okay so if I increasing here
624
00:29:13,600 --> 00:29:17,500
and I put one that is going to be randomized
625
00:29:17,500 --> 00:29:20,033
and that this is starting to help
626
00:29:20,300 --> 00:29:22,166
of course he's not perfect
627
00:29:22,266 --> 00:29:23,699
but he's starting to help
628
00:29:24,266 --> 00:29:27,799
then another thing that can help
629
00:29:27,966 --> 00:29:31,799
is the use of the stochastic samples
630
00:29:32,000 --> 00:29:33,800
this torcastic sample
631
00:29:33,800 --> 00:29:36,700
is going to render more information
632
00:29:36,700 --> 00:29:39,233
where you can see it
633
00:29:39,333 --> 00:29:42,466
and less information in the areas that are not
634
00:29:42,533 --> 00:29:43,766
so important
635
00:29:43,766 --> 00:29:47,266
and we are talking in here about perception
636
00:29:47,400 --> 00:29:51,800
so nuke is deciding where to use more information
637
00:29:51,800 --> 00:29:54,233
where the information is more dense
638
00:29:54,400 --> 00:29:57,333
and then applying less information
639
00:29:57,333 --> 00:30:02,666
that is going to create an effect of noise
640
00:30:02,766 --> 00:30:06,066
because it's going to create clusters of information
641
00:30:06,066 --> 00:30:09,799
so we are going to increase by zero
642
00:30:09,800 --> 00:30:12,633
you are getting just nothing
643
00:30:12,766 --> 00:30:14,666
but as soon as you put one
644
00:30:15,700 --> 00:30:18,500
you are going to start getting information in there
645
00:30:18,766 --> 00:30:20,066
one is very
646
00:30:20,066 --> 00:30:21,099
very very low
647
00:30:21,100 --> 00:30:22,566
so in this case
648
00:30:22,566 --> 00:30:23,999
as you can see here
649
00:30:24,000 --> 00:30:26,800
I have a good density of points
650
00:30:26,800 --> 00:30:28,400
and then in here
651
00:30:28,866 --> 00:30:29,999
that's not much
652
00:30:30,000 --> 00:30:31,733
see that there is nothing in there
653
00:30:31,733 --> 00:30:33,166
it's just valued zero
654
00:30:33,166 --> 00:30:34,466
okay of course
655
00:30:34,466 --> 00:30:36,299
if you are at this distance
656
00:30:36,400 --> 00:30:39,200
probably you are going to be more or less fine
657
00:30:39,333 --> 00:30:42,666
but it's not a good idea to use just one
658
00:30:42,866 --> 00:30:44,466
so we are going to increase
659
00:30:45,666 --> 00:30:49,499
up to the point that we can just get things to work
660
00:30:50,000 --> 00:30:51,000
good for us
661
00:30:52,533 --> 00:30:54,566
don't get confused
662
00:30:54,566 --> 00:30:58,266
this is not grain or noise
663
00:30:58,266 --> 00:31:02,033
this is that the result of getting the information
664
00:31:02,066 --> 00:31:06,299
group in clusters of a perception okay
665
00:31:06,300 --> 00:31:08,666
so with this difference is
666
00:31:08,666 --> 00:31:11,166
is going to take way less
667
00:31:11,333 --> 00:31:14,766
of course is related with the definition of my image
668
00:31:14,766 --> 00:31:16,266
if you have a very
669
00:31:16,266 --> 00:31:17,799
very noisy image
670
00:31:17,800 --> 00:31:22,333
that is going to create this kind of noisy pattern
671
00:31:22,333 --> 00:31:27,066
okay but in terms of the quality
672
00:31:27,100 --> 00:31:30,366
you need to get the final result of your image
673
00:31:30,366 --> 00:31:34,399
and then balance the stochastic samples as well
674
00:31:34,400 --> 00:31:36,500
okay so pray
675
00:31:38,000 --> 00:31:40,200
you get a good balance in there
676
00:31:40,700 --> 00:31:45,366
okay then once you have your good balance in there
677
00:31:45,400 --> 00:31:48,800
just try to apply the sample diameters
678
00:31:48,800 --> 00:31:53,133
the sample diameter is related with the antialias
679
00:31:53,133 --> 00:31:57,233
so is how big is the circle of pixels
680
00:31:57,300 --> 00:32:00,266
just to create the amount of antialia
681
00:32:00,266 --> 00:32:03,033
so the higher this is going to be
682
00:32:03,133 --> 00:32:06,499
the more accurate antialia as you are going to get
683
00:32:06,666 --> 00:32:10,199
that is for very precise calculation
684
00:32:10,200 --> 00:32:10,766
in this case
685
00:32:10,766 --> 00:32:12,799
because the motion blur is so high
686
00:32:13,066 --> 00:32:14,866
then you are not going to get
687
00:32:14,866 --> 00:32:17,866
that increase in the quality okay
688
00:32:18,333 --> 00:32:19,533
the focus diameter
689
00:32:19,533 --> 00:32:22,099
of course is related with the death of feel
690
00:32:22,100 --> 00:32:24,300
of the camera that you set up
691
00:32:24,300 --> 00:32:26,600
so I'm not going to talk about this
692
00:32:26,600 --> 00:32:30,966
but this is going to be the effect of the set blur
693
00:32:31,133 --> 00:32:32,533
done in camera
694
00:32:32,533 --> 00:32:34,266
so as you can imagine
695
00:32:34,266 --> 00:32:36,866
that said blur is just a combination
696
00:32:36,866 --> 00:32:38,233
as the vector blur
697
00:32:38,566 --> 00:32:41,066
but if you create the difocus
698
00:32:41,266 --> 00:32:43,066
directly from the calendar
699
00:32:43,066 --> 00:32:45,499
is going to work way better
700
00:32:45,566 --> 00:32:49,833
okay so the focus diameter is just going to define
701
00:32:50,300 --> 00:32:56,500
the amount of distribution of your samples okay
702
00:32:57,266 --> 00:32:58,699
good so now
703
00:32:58,700 --> 00:33:00,766
once I have my setup
704
00:33:00,900 --> 00:33:02,200
I can increase
705
00:33:04,133 --> 00:33:05,466
and see now
706
00:33:05,733 --> 00:33:08,199
by using 50 samples
707
00:33:08,600 --> 00:33:13,200
probably I'm going to get a quite decent result
708
00:33:13,900 --> 00:33:16,900
and by combining the stochastic samples
709
00:33:16,966 --> 00:33:19,066
and the amount of samples
710
00:33:19,200 --> 00:33:23,766
and reducing the time of the render by a huge fraction
711
00:33:23,766 --> 00:33:26,899
even if I was using the hundred samples
712
00:33:26,966 --> 00:33:28,999
that I was using initially
713
00:33:29,200 --> 00:33:30,966
the stocastic samples
714
00:33:30,966 --> 00:33:32,999
are going to reduce the time of the render
715
00:33:33,000 --> 00:33:36,066
because he's not using the full precision
716
00:33:36,066 --> 00:33:40,299
of every single pixel for blending the samples
717
00:33:40,300 --> 00:33:44,800
okay so the stocastic sample can be a very good asset
718
00:33:44,800 --> 00:33:47,866
to reduce the render time okay
719
00:33:48,366 --> 00:33:51,033
another thing that is interesting
720
00:33:51,200 --> 00:33:54,066
when we are working with the scanline render
721
00:33:54,266 --> 00:33:58,566
is to use your channels wisely
722
00:33:58,766 --> 00:34:00,566
so in this case
723
00:34:01,366 --> 00:34:01,999
for instance
724
00:34:02,000 --> 00:34:05,100
I'm creating my motion blur
725
00:34:05,200 --> 00:34:08,333
directly from my skyline render
726
00:34:08,333 --> 00:34:10,266
which means that probably
727
00:34:10,266 --> 00:34:16,566
I don't need my UV representing the motion vectors okay
728
00:34:16,600 --> 00:34:18,000
but in here
729
00:34:18,000 --> 00:34:18,966
by the fault
730
00:34:18,966 --> 00:34:23,633
this calendar is providing me with the motion vectors
731
00:34:23,666 --> 00:34:25,766
so what you want
732
00:34:25,766 --> 00:34:28,899
is to render just the information you need
733
00:34:29,000 --> 00:34:31,866
so here you go to this shader
734
00:34:32,000 --> 00:34:35,466
and then motion vectors off
735
00:34:36,366 --> 00:34:37,399
as you can see
736
00:34:37,400 --> 00:34:41,633
here you are rendering to channel less
737
00:34:42,733 --> 00:34:44,499
okay that's good
738
00:34:44,500 --> 00:34:45,700
and then you can say okay
739
00:34:45,700 --> 00:34:49,500
and what if I don't want my channel
740
00:34:49,500 --> 00:34:50,700
because in here
741
00:34:50,700 --> 00:34:54,900
I'm rendering also the death information well
742
00:34:55,300 --> 00:34:57,600
there is no way you can remove that channel
743
00:34:57,600 --> 00:35:01,500
even if you remove the set buffer or the transparency
744
00:35:01,533 --> 00:35:03,133
that's not related
745
00:35:03,133 --> 00:35:08,366
okay so if you want to get rid of that channel
746
00:35:08,733 --> 00:35:10,233
you just go in here
747
00:35:12,066 --> 00:35:18,633
remove and then you can either remove the deaf
748
00:35:19,100 --> 00:35:23,666
okay so is bypassing the other channels that's away
749
00:35:23,700 --> 00:35:25,300
or if you prefer
750
00:35:25,300 --> 00:35:27,800
you keep just the RGBA
751
00:35:28,066 --> 00:35:30,233
okay so he's going to discard
752
00:35:30,266 --> 00:35:33,166
any information that you don't need in there
753
00:35:33,166 --> 00:35:37,566
so that is going to create a very clean stream
754
00:35:37,566 --> 00:35:42,033
lining here with just the channels you really want
755
00:35:42,100 --> 00:35:45,500
okay and now we are going to discuss the boundary box
756
00:35:46,166 --> 00:35:49,233
in relation with the lens distortion
757
00:35:49,533 --> 00:35:52,699
so I have this image
758
00:35:53,133 --> 00:35:55,733
with the camera that I provided you
759
00:35:55,733 --> 00:35:56,733
and then in here
760
00:35:56,733 --> 00:36:00,666
I have the render of my cones okay
761
00:36:00,966 --> 00:36:04,799
so those cones are representing
762
00:36:04,800 --> 00:36:07,900
certain areas of the 3D space okay
763
00:36:08,800 --> 00:36:09,933
but in here
764
00:36:09,933 --> 00:36:13,766
in order to match my virtual image
765
00:36:14,100 --> 00:36:16,466
with the original plate
766
00:36:16,933 --> 00:36:20,866
I need to get my lenses torsion align
767
00:36:21,366 --> 00:36:25,966
and let me underline this point
768
00:36:26,200 --> 00:36:31,266
you never touch the original plate
769
00:36:31,666 --> 00:36:35,566
by using the and distort for compositing reasons
770
00:36:35,600 --> 00:36:37,366
you use the and distortion
771
00:36:37,366 --> 00:36:39,799
just for calculating your camera
772
00:36:39,800 --> 00:36:42,700
when you are applying the much more process
773
00:36:42,700 --> 00:36:46,066
but never never use this
774
00:36:46,066 --> 00:36:48,366
which is from the calendar
775
00:36:48,700 --> 00:36:51,333
then you apply the and distortion to the image
776
00:36:51,333 --> 00:36:54,499
and then put both together because
777
00:36:54,500 --> 00:36:55,966
yeah it's true
778
00:36:56,133 --> 00:36:59,699
they are both align in terms of lenses torsion
779
00:36:59,766 --> 00:37:02,199
but you are breaking in here
780
00:37:02,200 --> 00:37:05,800
one of the principles of the capacity
781
00:37:05,800 --> 00:37:07,966
which is you are touching
782
00:37:08,133 --> 00:37:10,399
something that is not mean to be touched
783
00:37:10,400 --> 00:37:12,200
which is the background plate
784
00:37:12,266 --> 00:37:14,333
if you don't have a particular task
785
00:37:14,333 --> 00:37:16,866
that requires that be touched
786
00:37:16,866 --> 00:37:20,899
so you are losing a huge amount of quality
787
00:37:21,900 --> 00:37:23,533
with the lens distortion
788
00:37:23,533 --> 00:37:25,799
which is applying a filter
789
00:37:25,866 --> 00:37:28,699
okay so never do that
790
00:37:28,766 --> 00:37:31,433
what you want is just to apply
791
00:37:31,666 --> 00:37:34,299
the lens distortion in here
792
00:37:34,300 --> 00:37:35,200
just to read
793
00:37:35,200 --> 00:37:36,966
distort your CG
794
00:37:37,266 --> 00:37:40,099
in order to match the original plate
795
00:37:40,166 --> 00:37:41,766
with the distortion
796
00:37:41,900 --> 00:37:44,766
um this is happening always
797
00:37:44,766 --> 00:37:46,999
there is no exception
798
00:37:47,200 --> 00:37:49,000
all cameras in the world
799
00:37:49,000 --> 00:37:53,200
have a certain amount of lens distortion
800
00:37:53,366 --> 00:37:54,133
some of them
801
00:37:54,133 --> 00:37:55,166
they have more
802
00:37:55,166 --> 00:37:57,466
some of them have less
803
00:37:57,666 --> 00:38:00,333
but there is always some
804
00:38:00,333 --> 00:38:03,399
okay so if I apply this
805
00:38:04,800 --> 00:38:08,900
to the process I did before with the stochastic samples
806
00:38:08,900 --> 00:38:11,466
okay you will see that now
807
00:38:11,533 --> 00:38:15,966
because of the filter I'm using here see that
808
00:38:15,966 --> 00:38:20,766
you are getting a bit of the reduction of that crane
809
00:38:20,866 --> 00:38:25,966
so you can play back with the stocastic samples
810
00:38:26,100 --> 00:38:27,300
and the filtering here
811
00:38:27,300 --> 00:38:29,566
so if I use something like the notch
812
00:38:29,733 --> 00:38:31,899
see that now is reducing
813
00:38:32,666 --> 00:38:35,966
because he's blending those samples okay
814
00:38:36,133 --> 00:38:39,099
so this is something that you can play with
815
00:38:39,100 --> 00:38:40,766
in case you need it
816
00:38:41,133 --> 00:38:43,299
so let's get back in here
817
00:38:43,300 --> 00:38:46,766
so I apply my lens distortion to this
818
00:38:46,766 --> 00:38:49,899
and what is this
819
00:38:50,366 --> 00:38:53,433
well this is totally normal
820
00:38:53,666 --> 00:38:56,766
is because when you are applying
821
00:38:58,300 --> 00:39:00,600
the lens distortion to your image
822
00:39:01,300 --> 00:39:02,766
you are doing this
823
00:39:02,933 --> 00:39:05,966
you are and distorting your image
824
00:39:06,166 --> 00:39:11,033
by taking the corners and stretching them out
825
00:39:11,133 --> 00:39:16,366
okay so the image seems to be enlarging
826
00:39:16,366 --> 00:39:23,066
I'm going to show you in here by using a black outside
827
00:39:24,866 --> 00:39:29,766
okay and then in here a note that I have in my toolbox
828
00:39:29,766 --> 00:39:32,066
which is the box to format
829
00:39:32,366 --> 00:39:37,266
so can you see what's happening in the edges
830
00:39:37,733 --> 00:39:42,299
the image is getting a stretch in the corners
831
00:39:43,366 --> 00:39:46,599
so if I change from there
832
00:39:48,066 --> 00:39:50,633
see that the image is actually looking like
833
00:39:50,666 --> 00:39:52,766
the center is getting back
834
00:39:52,766 --> 00:39:56,099
but it's actually the other way around is the corners
835
00:39:57,566 --> 00:40:02,266
so this is a non linear transformation of the image
836
00:40:02,266 --> 00:40:09,066
okay but what is actually happening is that this image
837
00:40:09,100 --> 00:40:11,133
that is getting outside
838
00:40:11,133 --> 00:40:13,899
because we are stretching the corners
839
00:40:13,966 --> 00:40:17,099
then when you apply redistort
840
00:40:20,500 --> 00:40:25,200
those areas that are outside are getting back inside
841
00:40:25,333 --> 00:40:29,499
okay but what if I apply this
842
00:40:29,566 --> 00:40:34,166
and then I crop my edges so
843
00:40:35,466 --> 00:40:36,433
I'm going to
844
00:40:39,933 --> 00:40:42,999
cropping here okay
845
00:40:47,366 --> 00:40:48,566
see what's going on
846
00:40:49,200 --> 00:40:53,866
is because this crop is erasing all the information
847
00:40:53,866 --> 00:40:56,299
that was in those corners
848
00:40:56,300 --> 00:40:57,700
outside the image
849
00:40:57,700 --> 00:40:59,800
so the first thing we cannot serve is
850
00:40:59,800 --> 00:41:02,766
the boundary box is expanding okay
851
00:41:03,133 --> 00:41:06,966
so what is happening to the scanline render
852
00:41:06,966 --> 00:41:08,466
is exactly that
853
00:41:09,866 --> 00:41:15,199
your render is just up to the format
854
00:41:15,200 --> 00:41:18,000
so when you apply the redistort
855
00:41:19,366 --> 00:41:23,866
those pixels that are supposed to be outside
856
00:41:23,866 --> 00:41:27,399
with the boundary box are not there anymore
857
00:41:27,700 --> 00:41:31,666
so this calendar is just rendering the format
858
00:41:31,766 --> 00:41:32,866
and then in here
859
00:41:32,866 --> 00:41:34,199
the format is getting inside
860
00:41:34,200 --> 00:41:38,566
so you are getting the regular issue of the expansion
861
00:41:38,666 --> 00:41:42,766
of the last pixel touching the boundary box
862
00:41:43,100 --> 00:41:45,766
okay so how can we fix that
863
00:41:45,766 --> 00:41:49,299
well there is a feature in the scan line render
864
00:41:50,133 --> 00:41:52,566
that is called the overscan
865
00:41:52,733 --> 00:41:57,033
so the overscan is just mean to expand
866
00:41:57,466 --> 00:42:02,566
whatever is not seen by the camera outside the format
867
00:42:02,600 --> 00:42:07,533
but is information that in a regular 3D virtual world
868
00:42:07,533 --> 00:42:08,699
we can apply
869
00:42:08,700 --> 00:42:11,100
so we can apply that
870
00:42:11,366 --> 00:42:16,366
and you can feel how the boundary box is expanding
871
00:42:16,366 --> 00:42:18,766
okay so when I move
872
00:42:18,766 --> 00:42:20,366
for instance in here
873
00:42:20,700 --> 00:42:23,266
I'm going to apply a transform
874
00:42:23,700 --> 00:42:24,700
just to move
875
00:42:25,066 --> 00:42:28,299
we are going to reveal a piece of the corn
876
00:42:28,800 --> 00:42:30,200
that is not seen
877
00:42:30,900 --> 00:42:34,766
as we did in the class talking about
878
00:42:36,100 --> 00:42:37,300
the boundary box
879
00:42:38,600 --> 00:42:41,066
see the cone is in there
880
00:42:41,100 --> 00:42:45,233
so if I restored that
881
00:42:45,933 --> 00:42:50,566
the information outside is going to be now insight
882
00:42:50,733 --> 00:42:54,533
due to the calculations of the lens distortion
883
00:42:54,533 --> 00:42:58,499
so now when I apply that to the image
884
00:42:59,400 --> 00:43:03,466
that is going to appear correctly
885
00:43:03,933 --> 00:43:05,066
the point is
886
00:43:05,333 --> 00:43:06,699
how they overcome
887
00:43:06,700 --> 00:43:10,400
should be well as big enough to cover that
888
00:43:10,400 --> 00:43:13,600
so it's something that you are going to operate by eye
889
00:43:13,866 --> 00:43:15,299
my advice is
890
00:43:15,333 --> 00:43:19,099
when you are going to calculate the oberscan
891
00:43:19,200 --> 00:43:21,800
go to the corners of the image
892
00:43:21,800 --> 00:43:24,200
so in case your cones
893
00:43:24,200 --> 00:43:26,833
are getting to the corner of the image
894
00:43:26,933 --> 00:43:30,099
that is the most critical piece of the image
895
00:43:30,100 --> 00:43:32,400
because remember that the lenses torsion
896
00:43:32,400 --> 00:43:36,300
is actually stretching from the corners of the image
897
00:43:36,300 --> 00:43:40,300
okay so there is going to be easier
898
00:43:40,466 --> 00:43:45,099
and then add a few pixels more in the overscan okay
899
00:43:45,700 --> 00:43:48,000
just to be sure you are not cropping
900
00:43:48,000 --> 00:43:49,466
but be careful
901
00:43:49,533 --> 00:43:52,199
because the higher is the overscan
902
00:43:52,366 --> 00:43:54,699
the more pixels you are going to render
903
00:43:54,700 --> 00:43:56,266
from your scan line render
904
00:43:56,266 --> 00:43:59,999
so you just want to keep the scan line render
905
00:44:00,000 --> 00:44:03,366
with less pixels to be render as possible
906
00:44:03,400 --> 00:44:06,300
without sacrifice the information you need
907
00:44:06,366 --> 00:44:10,566
then another thing that I used to do is I always
908
00:44:10,566 --> 00:44:12,033
when I use the merch
909
00:44:12,133 --> 00:44:16,366
put the information coming from the virtual world
910
00:44:16,766 --> 00:44:18,233
with the original plate
911
00:44:18,400 --> 00:44:20,800
usually I don't want to change
912
00:44:20,800 --> 00:44:23,700
that information in terms of transformation
913
00:44:23,700 --> 00:44:28,700
so what I do is just to simply crop my boundary box
914
00:44:28,866 --> 00:44:33,366
and limit that to the B input
915
00:44:33,366 --> 00:44:35,266
which is the original plate
916
00:44:35,266 --> 00:44:37,033
so in here B
917
00:44:37,066 --> 00:44:39,299
and now the boundary box
918
00:44:39,500 --> 00:44:44,100
is perfectly fit with all the information inside
919
00:44:44,466 --> 00:44:46,433
so I hope you enjoyed this
920
00:44:46,533 --> 00:44:51,299
set of tricks and methods for working with the 3D world
921
00:44:51,300 --> 00:44:54,400
okay so that's all for this week
922
00:44:54,533 --> 00:44:55,866
if you have any comments
923
00:44:55,866 --> 00:44:57,566
questions or suggestions
924
00:44:57,566 --> 00:44:59,899
you are very welcome to post in the forums
925
00:45:00,666 --> 00:45:02,033
until next time
926
00:45:02,066 --> 00:45:05,966
this has been Victor Perez for effects PhD
927
00:45:06,366 --> 00:45:07,566
thanks for watching
64901
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.