WEBVTT 1 00:00:23.650 --> 00:00:27.200 Hello, this is Juyoung Lim of Westworld 2 00:00:27.200 --> 00:00:31.100 In this lecture, I will talk about 3 00:00:31.100 --> 00:00:34.200 2D driving comp, an area of in-camera VFX, 4 00:00:34.200 --> 00:00:38.400 and through examples, how they are shot 5 00:00:38.400 --> 00:00:41.650 and used will also be explained 6 00:00:41.650 --> 00:00:45.350 Workflow of 2D driving comp technology 7 00:00:46.599 --> 00:00:50.449 I recommend that you learn about the most used 8 00:00:50.449 --> 00:00:53.049 ICVFX technology which is 2D driving comp 9 00:00:53.049 --> 00:00:55.000 about its general workflow 10 00:00:55.000 --> 00:00:58.050 and how it's prepared and shot 11 00:00:58.050 --> 00:01:01.680 through example cases 12 00:01:01.680 --> 00:01:06.959 First, let me remind you what 2D driving comp is 13 00:01:06.959 --> 00:01:12.680 It seems that driving scenes became a must in all drama series or movies 14 00:01:12.680 --> 00:01:17.030 This is because there are a lot of instances where explanation, dialog, and 15 00:01:17.030 --> 00:01:21.440 episodes take place while moving to somewhere 16 00:01:21.440 --> 00:01:24.490 Before, to film these scenes 17 00:01:24.490 --> 00:01:28.090 they brought cars into green studio 18 00:01:28.090 --> 00:01:32.690 or put cars on tow trucks 19 00:01:32.690 --> 00:01:34.720 to shoot these shots 20 00:01:34.720 --> 00:01:37.220 If you film in a green studio 21 00:01:37.220 --> 00:01:39.270 all works have to be done through compositing 22 00:01:39.270 --> 00:01:42.650 and the quality or problem of spill 23 00:01:42.650 --> 00:01:44.720 meaning green color showing off on other objects 24 00:01:44.720 --> 00:01:46.919 might become an issue 25 00:01:46.919 --> 00:01:49.419 Using tow trucks 26 00:01:49.419 --> 00:01:55.199 have less flexibility in filming time and safety 27 00:01:55.199 --> 00:02:01.149 But with 2D driving comp using LED wall 28 00:02:01.149 --> 00:02:02.949 time and spatial limitation 29 00:02:02.949 --> 00:02:06.499 and the advantages of driving comp 30 00:02:06.499 --> 00:02:10.650 were enhanced in the environment 31 00:02:10.650 --> 00:02:13.650 The biggest problem in driving comp is 32 00:02:13.650 --> 00:02:19.119 the absence of a source to use driving comp 33 00:02:19.119 --> 00:02:22.719 Driving comp needs a lot of source from 34 00:02:22.719 --> 00:02:26.839 a certain road, environment, or area 35 00:02:26.839 --> 00:02:32.139 Even a simple dialog needs to take the speed the car is going 36 00:02:32.139 --> 00:02:34.689 and put that into consideration 37 00:02:34.689 --> 00:02:37.720 which means that a lot of source is needed 38 00:02:37.720 --> 00:02:40.370 and that it requires a 39 00:02:40.370 --> 00:02:44.559 video shot in 360° 40 00:02:44.559 --> 00:02:46.359 You might think that 41 00:02:46.359 --> 00:02:50.409 a video shot with an action cam or 42 00:02:50.409 --> 00:02:56.320 a VR camera would do for the 360° video 43 00:02:56.320 --> 00:03:00.370 But images shot with VR cameras 44 00:03:00.370 --> 00:03:05.119 might have the total resolution of 10K or 12K 45 00:03:05.119 --> 00:03:07.619 but each video, 46 00:03:07.619 --> 00:03:11.019 that is, the area filmed by the camera 47 00:03:11.019 --> 00:03:14.119 might not even reach 2K 48 00:03:14.119 --> 00:03:19.719 For this reason, when making a source for 2D driving comp 49 00:03:19.719 --> 00:03:24.240 using an indoor camera is recommended 50 00:03:24.240 --> 00:03:28.590 So the most important thing of 2D driving comp is 51 00:03:28.590 --> 00:03:31.520 how to capture the source 52 00:03:31.520 --> 00:03:37.240 and also what spec that captured source has 53 00:03:37.240 --> 00:03:39.990 As you can see here, the captured source 54 00:03:39.990 --> 00:03:43.839 must have a data more than 10 bit 55 00:03:43.839 --> 00:03:46.989 and must be saved in the format of raw data 56 00:03:46.989 --> 00:03:51.000 which has the least damage 57 00:03:51.000 --> 00:03:53.350 And the capture speed 58 00:03:53.350 --> 00:03:56.650 means that if we film at 24FPS 59 00:03:56.650 --> 00:04:00.839 the captured source must also be shot at 24FPS 60 00:04:00.880 --> 00:04:04.680 and if the movie is shot at 48FPS 61 00:04:04.680 --> 00:04:08.919 the captured source must be captured at 48FPS as well 62 00:04:08.919 --> 00:04:11.919 Only that way there will be no empty frames 63 00:04:11.919 --> 00:04:15.360 and all frames will be perfectly captured 64 00:04:15.360 --> 00:04:18.510 So it is important at what level of resolution 65 00:04:18.510 --> 00:04:21.000 the provided camera films 66 00:04:21.000 --> 00:04:25.100 Most action cameras or VR cameras 67 00:04:25.100 --> 00:04:29.480 don't provide this high-end data format 68 00:04:29.480 --> 00:04:32.430 So it is inevitable to have the challenge of 69 00:04:32.430 --> 00:04:36.530 using many cameras and film at the same time 70 00:04:36.530 --> 00:04:39.279 and rig them 71 00:04:39.279 --> 00:04:41.029 I will give a more detailed explanation on 72 00:04:41.029 --> 00:04:43.869 capture frame rate 73 00:04:43.869 --> 00:04:46.769 Capture frame rate is what speed 74 00:04:46.769 --> 00:04:51.000 a scene is captured in 75 00:04:51.000 --> 00:04:53.750 If it's 24 frames in a second 76 00:04:53.750 --> 00:04:56.350 it is taking 24 photos in 1 second 77 00:04:56.350 --> 00:04:59.559 and making a 1 second long video 78 00:04:59.559 --> 00:05:04.259 The reason why this is important is we take this captured source 79 00:05:04.259 --> 00:05:08.239 and combine it with LED walls and film 80 00:05:08.239 --> 00:05:11.939 If a source filmed at 24FPS 81 00:05:11.939 --> 00:05:16.399 is converted to 48 FPS 82 00:05:16.440 --> 00:05:19.390 there won't be empty frames 83 00:05:19.390 --> 00:05:21.720 but there will be paused frames 84 00:05:21.720 --> 00:05:24.120 A finger would have moved 85 00:05:24.120 --> 00:05:29.279 but it wouldn't move in certain frames 86 00:05:29.279 --> 00:05:32.329 This might not be notifiable to human eye 87 00:05:32.329 --> 00:05:35.029 but it is inevitable that 88 00:05:35.029 --> 00:05:38.720 a pausing video is film if a 48FPS camera 89 00:05:38.720 --> 00:05:42.799 films simply a converting video 90 00:05:42.799 --> 00:05:45.349 So the captured frame 91 00:05:45.349 --> 00:05:47.949 must be identical or more 92 00:05:47.949 --> 00:05:51.499 than the frame rate that is 93 00:05:51.499 --> 00:05:55.320 used in the actual production 94 00:05:55.320 --> 00:05:57.470 This is more about how much 95 00:05:57.470 --> 00:06:00.970 the camera that is currently being used 96 00:06:00.970 --> 00:06:05.839 provides rather than about development 97 00:06:05.880 --> 00:06:08.959 The second thing is sync 98 00:06:08.959 --> 00:06:11.759 Sync here means 99 00:06:11.759 --> 00:06:14.709 if many cameras film at the same time 100 00:06:14.709 --> 00:06:19.440 that they must film the same scene on the same frame 101 00:06:19.440 --> 00:06:23.240 Same frame doesn't mean the same spot 102 00:06:23.240 --> 00:06:25.790 but that when I capture a scene 103 00:06:25.790 --> 00:06:30.519 another camera in another angle must capture the same scene in the same time axis 104 00:06:30.519 --> 00:06:33.319 To achieve this 105 00:06:33.319 --> 00:06:38.000 we need the sync system I mentioned in ICVFX 106 00:06:38.000 --> 00:06:40.450 But except for these major cameras 107 00:06:40.450 --> 00:06:44.679 there aren't much cameras that provide this sync system 108 00:06:44.679 --> 00:06:48.880 But there is no other way than using cameras that provide sync system 109 00:06:48.880 --> 00:06:52.930 and along with the sync system 110 00:06:52.930 --> 00:06:55.080 we also need a system that can send off 111 00:06:55.080 --> 00:06:58.040 a signal to start shooting 112 00:06:58.040 --> 00:07:02.140 If the shoot is done with 8, 9, or 11 cameras 113 00:07:02.140 --> 00:07:04.490 if the record buttons aren't pressed at the same time 114 00:07:04.490 --> 00:07:07.480 the sync will be off from the beginning 115 00:07:07.480 --> 00:07:11.760 so there needs to be a system that can check and control this 116 00:07:11.760 --> 00:07:14.060 The third thing is 117 00:07:14.060 --> 00:07:17.839 how to place the cameras 118 00:07:17.839 --> 00:07:20.789 Camera placement, as you saw just before, 119 00:07:20.789 --> 00:07:23.089 can be outside the car 120 00:07:23.089 --> 00:07:25.799 or it could be in a circle 121 00:07:25.799 --> 00:07:31.359 Which one is better depends on the project or filming method 122 00:07:31.359 --> 00:07:35.309 The circular plate method that we use 123 00:07:35.309 --> 00:07:39.659 is good for when we combine the images 124 00:07:39.659 --> 00:07:42.519 filmed in many cameras into a single image 125 00:07:42.519 --> 00:07:47.719 But since the cameras are connected to each other 126 00:07:47.719 --> 00:07:51.569 there are less space, and controlling 127 00:07:51.569 --> 00:07:54.959 the heat produced or the wind can be difficult 128 00:07:54.959 --> 00:07:57.559 There will also be the issue of 129 00:07:57.559 --> 00:08:00.459 how to attach this rig with 9 cameras 130 00:08:00.459 --> 00:08:05.160 to the car or to a certain other object 131 00:08:05.160 --> 00:08:09.610 Attaching cameras to the outside of a car one by one 132 00:08:09.610 --> 00:08:11.910 might be convenient for filming and attaching 133 00:08:11.910 --> 00:08:15.560 but for combining the filmed data into one 134 00:08:15.560 --> 00:08:18.480 it will require more labor and effort 135 00:08:18.480 --> 00:08:22.530 Because stitching must 136 00:08:22.530 --> 00:08:25.130 be centered around and combined into one based on one origin 137 00:08:25.130 --> 00:08:28.230 but each camera will have different origins 138 00:08:28.230 --> 00:08:31.720 so there needs a process of adjusting the origins 139 00:08:31.720 --> 00:08:36.720 What you're watching now is a video of when stitching is done, 140 00:08:36.720 --> 00:08:38.920 what human eyes see, what a normal player shows, 141 00:08:38.920 --> 00:08:42.320 and what a LED wall shows 142 00:08:42.359 --> 00:08:44.509 Another problems is that when stitching is done 143 00:08:44.509 --> 00:08:47.009 there is camera ghosting 144 00:08:47.009 --> 00:08:51.640 Camera ghosting is a problem that occurs from the angle of camera array 145 00:08:51.640 --> 00:08:53.790 Like the bus you see there 146 00:08:53.790 --> 00:08:58.240 if two cameras filmed that bus and stitched them 147 00:08:58.240 --> 00:09:01.740 there will be a slight difference in the angle 148 00:09:01.740 --> 00:09:04.549 which would create some awkwardess 149 00:09:04.599 --> 00:09:07.099 How to solve this problem 150 00:09:07.099 --> 00:09:11.249 or how you're going to hide this 151 00:09:11.249 --> 00:09:14.000 is another thing you must think about 152 00:09:14.000 --> 00:09:16.850 Once it is determined how to film the image plate 153 00:09:16.850 --> 00:09:20.159 scheduling for the project takes place 154 00:09:20.159 --> 00:09:22.809 The image you're seeing is a schedule table 155 00:09:22.809 --> 00:09:27.320 for a general 2D driving or ICVFX using image plate 156 00:09:27.320 --> 00:09:30.370 Unlike 3D ICVFX, 157 00:09:30.370 --> 00:09:33.520 filming step of asset preparation stage 158 00:09:33.520 --> 00:09:37.479 and testing step of LED volume is added 159 00:09:37.479 --> 00:09:43.529 From this, scheduling is done by calculating the time that will take to film 160 00:09:43.529 --> 00:09:48.640 and test unlike 3D ICVFX 161 00:09:48.679 --> 00:09:52.179 Examples of 2D driving comp 162 00:09:52.829 --> 00:09:55.729 Based on the things I just explained 163 00:09:55.729 --> 00:09:59.729 I will give a more detailed explanation 164 00:09:59.729 --> 00:10:03.679 using D.P. season 2's driving comp 165 00:10:03.679 --> 00:10:09.200 and train composition scene that we did 166 00:10:09.200 --> 00:10:12.850 D.P. season 2 is a Netflix project released last year 167 00:10:12.850 --> 00:10:17.039 and the shoot was during the end of 2022 168 00:10:17.039 --> 00:10:20.989 And after the shooting and some post-production processes 169 00:10:20.989 --> 00:10:24.640 it was released on July last year 170 00:10:24.640 --> 00:10:28.990 You can consider D.P. season 2 project is a project 171 00:10:28.990 --> 00:10:31.919 that goes through all steps of a driving comp 172 00:10:31.919 --> 00:10:36.019 It was the first case where normal car, 173 00:10:36.019 --> 00:10:41.200 several cars' interaction, and lastly a long train's driving comp shot 174 00:10:41.200 --> 00:10:45.750 were all tested and actually filmed 175 00:10:45.750 --> 00:10:49.359 to make the final plate 176 00:10:49.359 --> 00:10:54.109 Through this, you will be able to learn 177 00:10:54.109 --> 00:10:56.960 how sources must be captured, 178 00:10:56.960 --> 00:10:59.510 how to prepare the captured source, 179 00:10:59.510 --> 00:11:03.840 and what to do when the captured source isn't enough 180 00:11:03.840 --> 00:11:09.540 D.P. season 2 took about 3 months of preparation 181 00:11:09.540 --> 00:11:11.359 before the production 182 00:11:11.359 --> 00:11:13.409 It wasn't really that short 183 00:11:13.409 --> 00:11:17.840 and thankfully we had accumulated a decent amount of data 184 00:11:17.840 --> 00:11:20.690 on car driving scene capturing 185 00:11:20.690 --> 00:11:22.960 during another project we had done before D.P. season 2 186 00:11:22.960 --> 00:11:25.510 And we also had the equipment needed 187 00:11:25.510 --> 00:11:29.039 so the R&D of car scene was pretty much ready 188 00:11:29.039 --> 00:11:33.689 but the R&D for the train scene which was the main thing 189 00:11:33.689 --> 00:11:38.320 and solving technical issues spent a lot of the R&D cost 190 00:11:38.320 --> 00:11:44.640 Next was filming and source production for the empty parts 191 00:11:44.640 --> 00:11:48.960 So it was a project reaching towards the actual production 192 00:11:48.960 --> 00:11:52.910 What's different from 3D ICVFX I explained on the third lecture is 193 00:11:52.910 --> 00:11:58.320 is that 2D ICVFX only had a virtual production supervisor 194 00:11:58.320 --> 00:12:01.570 This was because the team in charge of VFX was 195 00:12:01.570 --> 00:12:03.840 a different team from us 196 00:12:03.840 --> 00:12:05.640 so it was another challenge 197 00:12:05.640 --> 00:12:10.159 to do virtual production working together with that team 198 00:12:10.159 --> 00:12:13.440 So we had a more departmentalized structure 199 00:12:13.440 --> 00:12:16.840 and we divided into teams that does 200 00:12:16.840 --> 00:12:19.690 virtual production and a team that 201 00:12:19.690 --> 00:12:22.880 processes sources and worked in a dual structure 202 00:12:22.880 --> 00:12:26.500 What we did first was storyboard analysis 203 00:12:26.500 --> 00:12:28.950 We actually only had 204 00:12:28.950 --> 00:12:32.719 the plan that we were going to film car and train scenes 205 00:12:32.719 --> 00:12:37.039 and didn't have anything planned for what scene to shoot for what episode 206 00:12:37.039 --> 00:12:40.080 So we analyzed the storyboard 207 00:12:40.080 --> 00:12:42.630 coming up with plans that 208 00:12:42.630 --> 00:12:46.080 we could film this kind of shots but not these ones 209 00:12:46.080 --> 00:12:48.430 and we scheduled to do 210 00:12:48.430 --> 00:12:51.520 a tech-viz based on the storyboard 211 00:12:51.520 --> 00:12:54.070 Of course there were some challenges 212 00:12:54.070 --> 00:13:00.400 but we proposed that we do as much driving comp as possible 213 00:13:00.400 --> 00:13:02.500 Once what scene would be shot 214 00:13:02.500 --> 00:13:05.119 through storyboard analysis was done 215 00:13:05.119 --> 00:13:08.479 we started talking about LED volume design 216 00:13:08.479 --> 00:13:10.479 Volume for the car scene 217 00:13:10.479 --> 00:13:12.879 was possible to cover 218 00:13:12.879 --> 00:13:15.919 with the volume used in Black Knight 219 00:13:15.919 --> 00:13:19.019 But the train scene 220 00:13:19.019 --> 00:13:25.280 would need way more than the one used in Black Knight 221 00:13:25.280 --> 00:13:28.380 So we used various tech-viz methods 222 00:13:28.380 --> 00:13:34.799 to test what would be able to cover for the most area 223 00:13:34.799 --> 00:13:38.199 How we could place the train 224 00:13:38.199 --> 00:13:41.440 to use the Black Knight LED volume 225 00:13:41.440 --> 00:13:47.340 If only one side of the train had the windows open 226 00:13:47.340 --> 00:13:50.400 if that could minimize the settings 227 00:13:50.400 --> 00:13:54.799 For car scenes, how the setting must be done depending on the size of the cars 228 00:13:54.799 --> 00:13:58.049 We took all these into consideration 229 00:13:58.049 --> 00:14:01.919 and searched for the final LED design volume 230 00:14:01.919 --> 00:14:03.519 What was good here was 231 00:14:03.519 --> 00:14:06.669 when we were doing this tech-viz 232 00:14:06.669 --> 00:14:10.960 we used a media engine instead of game engines 233 00:14:10.960 --> 00:14:14.110 Using the 3D data import function of the media engine, 234 00:14:14.110 --> 00:14:18.239 as you can see, we didn't only bring the LED wall 235 00:14:18.239 --> 00:14:21.389 but also the car that will be used 236 00:14:21.389 --> 00:14:24.189 and the mesh of the train were also brought in 237 00:14:24.189 --> 00:14:26.439 to calculate the size of the window 238 00:14:26.439 --> 00:14:30.239 and the glass to be used 239 00:14:30.239 --> 00:14:32.389 which is the function that helped us 240 00:14:32.389 --> 00:14:35.119 greatly when designing the LED volume 241 00:14:35.119 --> 00:14:37.440 We continuously tested various functions 242 00:14:37.440 --> 00:14:40.290 and had options like maybe a straight setting will be better 243 00:14:40.290 --> 00:14:42.790 or since the truck will be there 244 00:14:42.790 --> 00:14:45.280 if might be alright to keep the circular structure 245 00:14:45.280 --> 00:14:48.640 Especially how the source was screened was a very important issue 246 00:14:48.640 --> 00:14:50.690 Car sources usually 247 00:14:50.690 --> 00:14:53.679 screen a circular array 248 00:14:53.679 --> 00:14:57.929 but trains are longer so we wondered if 249 00:14:57.929 --> 00:15:00.329 it should be shot longer and wider or have a different array 250 00:15:00.329 --> 00:15:02.159 These things we also had to think about 251 00:15:02.159 --> 00:15:06.959 so we designed the LED volume thinking about 252 00:15:06.959 --> 00:15:08.659 how the train would look on screen 253 00:15:08.659 --> 00:15:10.559 if we put it parallelly 254 00:15:10.559 --> 00:15:12.799 and how the time axis would look 255 00:15:12.799 --> 00:15:15.549 So at the end, for the car scene, 256 00:15:15.549 --> 00:15:20.080 it was decided to use the original LED 257 00:15:20.080 --> 00:15:25.679 and for the train scene we decided to form a completely new LED volume 258 00:15:25.760 --> 00:15:30.510 So in the train most actions would take place in the aisle 259 00:15:30.510 --> 00:15:35.760 and the angle itself focuses on certain characters 260 00:15:35.760 --> 00:15:38.910 so we covered one side as much as possible 261 00:15:38.910 --> 00:15:42.559 and decided to block the other side where it reflects 262 00:15:42.559 --> 00:15:45.909 and where there is no LED in the middle 263 00:15:45.909 --> 00:15:49.759 we used curtains or people to 264 00:15:49.759 --> 00:15:51.919 cover that part up 265 00:15:51.919 --> 00:15:53.869 This final design 266 00:15:53.869 --> 00:15:56.369 would have of course been better 267 00:15:56.369 --> 00:15:59.919 if we could have LED on both sides 268 00:15:59.919 --> 00:16:03.469 but to give the best effect within the given resource 269 00:16:03.469 --> 00:16:05.679 this was what we settled on 270 00:16:05.679 --> 00:16:08.779 And with this design 271 00:16:08.779 --> 00:16:11.679 we did the pre-viz 272 00:16:11.679 --> 00:16:13.679 What cource would play 273 00:16:13.679 --> 00:16:16.279 and how they would be placed 274 00:16:16.279 --> 00:16:19.359 How much it actually covers 275 00:16:19.359 --> 00:16:21.609 and what windows need curtains over them 276 00:16:21.609 --> 00:16:24.459 and where actors must be 277 00:16:24.459 --> 00:16:27.280 These were all tested 278 00:16:27.280 --> 00:16:29.630 Once this test was done 279 00:16:29.630 --> 00:16:32.400 an actual train was placed 280 00:16:32.400 --> 00:16:34.500 I had never known 281 00:16:34.500 --> 00:16:37.679 that there were trucks to move trains 282 00:16:37.679 --> 00:16:41.359 and that trains can be lifted up with craines 283 00:16:41.359 --> 00:16:44.259 I also got to know that train wheels are also detachable 284 00:16:44.259 --> 00:16:46.459 Anyways we started with 285 00:16:46.459 --> 00:16:49.840 bringing a giant train into the studio 286 00:16:49.840 --> 00:16:53.190 The reason why we brought the train in first before building the LED walls 287 00:16:53.190 --> 00:16:57.359 was because we needed the train to get the right angle for the LED wall to be 288 00:16:57.359 --> 00:17:00.960 and because there was no way for the train to be brought in once the LED wall would be built 289 00:17:00.960 --> 00:17:03.110 So we started with bringing 290 00:17:03.110 --> 00:17:06.560 the train into the set 291 00:17:06.560 --> 00:17:09.310 After that we started building the LED 292 00:17:09.310 --> 00:17:11.610 and it followed the design 293 00:17:11.610 --> 00:17:14.000 that we had made 294 00:17:14.000 --> 00:17:17.550 It took about a week to set the LED up 295 00:17:17.550 --> 00:17:22.050 and after they were built, it took another week to 296 00:17:22.050 --> 00:17:25.839 synchronize and test the LED 297 00:17:25.839 --> 00:17:27.439 While the LED volume was being built 298 00:17:27.439 --> 00:17:30.319 we did tech-viz in the meantime 299 00:17:30.319 --> 00:17:33.019 The tech-viz combined and used 300 00:17:33.019 --> 00:17:35.019 media server and 301 00:17:35.019 --> 00:17:38.160 real time camera tracking technology used in real time preview 302 00:17:38.160 --> 00:17:40.710 The yellow area in the video 303 00:17:40.710 --> 00:17:44.660 are the parts excluding the actual chairs in the train that can move 304 00:17:44.660 --> 00:17:47.920 and the green area are the space to the outer wall of the train 305 00:17:47.920 --> 00:17:52.470 We tested with an actual camera 306 00:17:52.470 --> 00:17:55.860 how much within this space could be covered 307 00:17:55.860 --> 00:17:58.060 with the built LED and for the parts 308 00:17:58.060 --> 00:18:00.910 that can't be, how we would need to 309 00:18:00.910 --> 00:18:02.239 change the camera angle 310 00:18:02.239 --> 00:18:05.289 This was the first case to use 311 00:18:05.289 --> 00:18:07.520 media server and real time camera tracking 312 00:18:07.520 --> 00:18:10.870 and through this, we could check that 313 00:18:10.870 --> 00:18:13.680 not only game engines but also 314 00:18:13.680 --> 00:18:16.530 media servers and 2D plates can make use of 315 00:18:16.530 --> 00:18:21.760 real time camera tracking technology, that real time previews have 316 00:18:21.770 --> 00:18:28.239 Next was testing how to produce sources using the train 317 00:18:28.239 --> 00:18:35.040 We did various tests like if the 180° video just needs to be 318 00:18:35.040 --> 00:18:41.090 filmed from windows on one side of the train or if it's better to film from different windows 319 00:18:41.090 --> 00:18:43.040 with some time difference 320 00:18:43.040 --> 00:18:46.560 What was used in this test was game engine technology 321 00:18:46.560 --> 00:18:51.439 We made a virtual area identical to the train in the game engine 322 00:18:51.439 --> 00:18:55.839 and arrayed cameras that simulate real physical cameras 323 00:18:55.839 --> 00:19:00.640 and then checked how it would look 324 00:19:00.640 --> 00:19:02.240 when they were composed at the end 325 00:19:02.240 --> 00:19:04.560 with the videos filmed while the camera was moving 326 00:19:04.560 --> 00:19:07.599 Even though there were plenty of windows in the train 327 00:19:07.599 --> 00:19:10.549 we were able to confirm that we must stitch video 328 00:19:10.549 --> 00:19:13.359 taken from one spot 329 00:19:13.359 --> 00:19:16.640 but since this caused too much distortion 330 00:19:16.640 --> 00:19:19.390 we thought of ways to decrease 331 00:19:19.390 --> 00:19:21.119 this distortion 332 00:19:21.119 --> 00:19:24.119 Based on these various data 333 00:19:24.119 --> 00:19:27.760 we started gathering real production source data 334 00:19:27.760 --> 00:19:30.610 There were cars prepared for 335 00:19:30.610 --> 00:19:33.199 360° shots 336 00:19:33.199 --> 00:19:36.349 but the majority of cars we would film 337 00:19:36.349 --> 00:19:41.280 included sports cars, trucks, and SUVs 338 00:19:41.280 --> 00:19:45.180 SUVs or trucks could use sources 339 00:19:45.180 --> 00:19:48.640 filmed from normal cars with 360° filming 340 00:19:48.719 --> 00:19:53.359 but sports cars, even if it was filmed from itself, 341 00:19:53.359 --> 00:19:57.159 when we calculated the height of the camera rig, 342 00:19:57.159 --> 00:20:00.309 the height that is captured in 343 00:20:00.309 --> 00:20:01.952 didn't match 344 00:20:01.952 --> 00:20:06.752 So for sports cars we had to split them into two 180° shots 345 00:20:06.752 --> 00:20:11.322 and put cameras on the tow trucks 346 00:20:11.322 --> 00:20:13.622 And as we were filming 347 00:20:13.622 --> 00:20:17.922 we had to race on dirt road 348 00:20:17.922 --> 00:20:20.822 instead of normal paved roads 349 00:20:20.822 --> 00:20:22.672 And being on those unpaved surfaces 350 00:20:22.672 --> 00:20:26.002 cameras would shake too much 351 00:20:26.002 --> 00:20:30.260 and the shaking wasn't in the same lane 352 00:20:30.260 --> 00:20:35.490 It even caused the camera array to be break down 353 00:20:35.490 --> 00:20:38.390 so it would not match each other when doing the stitching 354 00:20:38.390 --> 00:20:41.519 or more camera ghosting would happen 355 00:20:41.519 --> 00:20:44.130 So to help this 356 00:20:44.130 --> 00:20:46.379 we added an equipment called damper 357 00:20:46.379 --> 00:20:49.959 so that it would absorb most of the shock 358 00:20:49.959 --> 00:20:55.909 With this updated rig, the car plate was filmed 359 00:20:55.919 --> 00:21:00.439 What you're seeing is the videos took with the updated rig and set 360 00:21:00.439 --> 00:21:02.520 As we were on those dirt roads 361 00:21:02.520 --> 00:21:05.760 without the rig that we had made that had dampers 362 00:21:05.760 --> 00:21:07.330 the camera would keep falling apart 363 00:21:07.330 --> 00:21:10.760 and stitching wouldn't be possible 364 00:21:10.760 --> 00:21:12.560 but since we prepared pretty well 365 00:21:12.560 --> 00:21:14.720 the shooting was done safely 366 00:21:14.720 --> 00:21:18.440 So we took car source scenes from a total of 3 areas 367 00:21:18.440 --> 00:21:20.590 They mostly had backgrounds of 368 00:21:20.590 --> 00:21:22.830 demilitarized zones connected to North Korea 369 00:21:22.830 --> 00:21:26.830 so we shot these shots in dirt roads, 370 00:21:26.830 --> 00:21:29.639 bushy areas, and where there are mountains 371 00:21:29.639 --> 00:21:32.160 I assume that D.P. season to would be 372 00:21:32.160 --> 00:21:35.360 the first in Korea to film in these 373 00:21:35.360 --> 00:21:38.000 rough places for driving comp 374 00:21:38.000 --> 00:21:41.040 Usually it is done in Gangnam-daero or Gwanghwamun 375 00:21:41.040 --> 00:21:44.279 where the night view is pretty 376 00:21:44.279 --> 00:21:46.839 but since this one had the theme of the military 377 00:21:46.839 --> 00:21:51.639 we captured a lot of sources from these unpaved roads 378 00:21:51.639 --> 00:21:55.150 Like this, car sources were pretty much ready 379 00:21:55.150 --> 00:21:56.100 and next 380 00:21:56.100 --> 00:21:59.919 we had to gather sources for the train 381 00:21:59.919 --> 00:22:04.480 There wasn't a way to attach cameras to the exterior of the train 382 00:22:04.480 --> 00:22:07.559 Even Mugunghwa-ho, 383 00:22:07.559 --> 00:22:11.520 we couldn't film from outside like we could before 384 00:22:11.520 --> 00:22:13.119 We had to film it only from the inside 385 00:22:13.119 --> 00:22:17.080 Then how to we attain source from the inside? 386 00:22:17.080 --> 00:22:19.199 This was the challenge 387 00:22:19.199 --> 00:22:21.240 We were going to film through the window 388 00:22:21.240 --> 00:22:23.119 but how we would rig to the window 389 00:22:23.119 --> 00:22:26.199 and how we would minimize the artifacts 390 00:22:26.199 --> 00:22:29.679 were tested 391 00:22:29.679 --> 00:22:33.080 Also, what was important in the train scene was the schedule 392 00:22:33.080 --> 00:22:34.520 We would have to get on the train 393 00:22:34.520 --> 00:22:37.839 For car scenes, for car sources 394 00:22:37.839 --> 00:22:42.240 we just would have to tell the driver 395 00:22:42.240 --> 00:22:45.000 Please go that way from the Gwanghwamun crossroad 396 00:22:45.000 --> 00:22:47.479 Go 50km/h in Gangnam-daero please 397 00:22:47.479 --> 00:22:48.919 Stop here then go there please 398 00:22:48.919 --> 00:22:50.800 All these can be controlled 399 00:22:50.800 --> 00:22:52.080 but for trains, that wasn't the case 400 00:22:52.080 --> 00:22:55.479 If it departs from Seoul station, it could only 401 00:22:55.479 --> 00:22:58.000 go to our destination in a constant speed 402 00:22:58.000 --> 00:23:02.600 So then what was important was we had to have a detailed plan 403 00:23:02.600 --> 00:23:04.250 on how to prepare at what timing 404 00:23:04.250 --> 00:23:06.919 and film at what timing 405 00:23:06.919 --> 00:23:10.279 As you can see we get on the train at Seoul station 406 00:23:10.279 --> 00:23:13.559 and rig the camera and prepare 407 00:23:13.559 --> 00:23:15.720 Wipe the windows 408 00:23:15.720 --> 00:23:19.240 Get of where, all these were prepared 409 00:23:19.240 --> 00:23:23.960 so that we could gather the sources we want following this schedule 410 00:23:23.960 --> 00:23:26.000 And to make this schedule 411 00:23:26.000 --> 00:23:30.399 we took trains from Seoul Station to Daejeon about 3 times 412 00:23:30.399 --> 00:23:32.839 to check the actual time 413 00:23:32.839 --> 00:23:35.600 About this much after we see country sites 414 00:23:35.600 --> 00:23:38.000 About this much after we arrive at Cheonan-Asan Station 415 00:23:38.000 --> 00:23:40.800 About this much after we're past Cheonan-Asan Station and see cities 416 00:23:40.800 --> 00:23:43.360 and then lastly we enter Daejeon this way 417 00:23:43.360 --> 00:23:47.720 We checked all these and made the schedule 418 00:23:47.720 --> 00:23:49.639 Next was camera rigging 419 00:23:49.639 --> 00:23:53.119 We tested for rigs to stick to windows 420 00:23:53.119 --> 00:23:56.759 This again couldn't be tested on the actual shooting 421 00:23:56.759 --> 00:24:00.000 so we took train from Seoul to Busan 422 00:24:00.000 --> 00:24:03.639 and took it on and off from the windows 423 00:24:03.639 --> 00:24:06.160 It stays on for about this long 424 00:24:06.160 --> 00:24:07.839 It takes about this long 425 00:24:07.839 --> 00:24:10.800 we checked these things 426 00:24:10.800 --> 00:24:12.100 And the most important challenge was 427 00:24:12.100 --> 00:24:15.520 once we did the filming and got the data 428 00:24:15.520 --> 00:24:18.079 we had the issue that we had to keep the windows clean 429 00:24:18.079 --> 00:24:20.559 We initially had thought that 430 00:24:20.559 --> 00:24:26.920 if the sun was out, the outside would be captured with the camera 431 00:24:26.920 --> 00:24:31.040 but since we had the focus as open as possible 432 00:24:31.040 --> 00:24:34.679 all the dirt on the window were noticeable 433 00:24:34.679 --> 00:24:39.200 So finding the spot with the cleanest window was the first issue 434 00:24:39.200 --> 00:24:45.000 and the second was to keep it as clean as possible 435 00:24:45.000 --> 00:24:48.160 So finding solution for the lighting on the window 436 00:24:48.160 --> 00:24:50.760 and keeping the window clean 437 00:24:50.760 --> 00:24:53.799 For that we even got a person to wipe the window 438 00:24:53.799 --> 00:24:56.279 that would get off on each station 439 00:24:56.279 --> 00:24:58.920 to wipe the window and then rig the camera back again 440 00:24:58.920 --> 00:25:01.399 Continuing this process 441 00:25:01.399 --> 00:25:04.679 we did the source capturing 442 00:25:04.679 --> 00:25:07.440 After that, with the captured data 443 00:25:07.440 --> 00:25:10.440 we finally did the stitching 444 00:25:10.440 --> 00:25:15.240 From here, we used compositing software and matching of data 445 00:25:15.240 --> 00:25:16.880 to see how cameras were arrayed 446 00:25:16.880 --> 00:25:18.799 and how it looks in that array 447 00:25:18.799 --> 00:25:22.160 How to fill in the empty areas 448 00:25:22.160 --> 00:25:24.760 and minimizing camera ghosting 449 00:25:24.760 --> 00:25:28.720 We found the optimal point and did the compositing 450 00:25:28.720 --> 00:25:32.679 Here, finding the optimal value was a challenge 451 00:25:32.679 --> 00:25:37.320 but the source we captured was about 25 minutes in length 452 00:25:37.330 --> 00:25:39.920 They were all captured in 24FPS 453 00:25:39.920 --> 00:25:42.160 so the frame amount was a lot 454 00:25:42.160 --> 00:25:46.750 and rendering and combining frames 455 00:25:46.750 --> 00:25:49.559 also took a lot of resource 456 00:25:49.559 --> 00:25:54.600 And for D.P. season 2, what's interesting is 457 00:25:54.600 --> 00:25:57.239 we also made environments using game engines 458 00:25:57.239 --> 00:25:58.640 The reason was 459 00:25:58.640 --> 00:26:00.559 one of the scenes the director wanted was 460 00:26:00.559 --> 00:26:03.119 going quickly from a bright place to a dark place 461 00:26:03.119 --> 00:26:05.440 He wanted a tunnel scene 462 00:26:05.440 --> 00:26:08.799 But there wan't a long enough tunnel in Korea 463 00:26:08.799 --> 00:26:11.640 Especially in the route from Seould to Daejeon 464 00:26:11.640 --> 00:26:13.920 there wasn't any long tunnel 465 00:26:13.920 --> 00:26:18.679 and it was pretty much impossible to adjust the iris diaphragm and film inside a tunnel 466 00:26:18.679 --> 00:26:20.320 because it was too dark 467 00:26:20.320 --> 00:26:23.880 So we made the tunnel section in 3D 468 00:26:23.880 --> 00:26:27.359 and we proposed that we combine that section 469 00:26:27.359 --> 00:26:30.679 with 2D captured images and play them together 470 00:26:30.679 --> 00:26:33.640 For that, we made a virtual tunnel 471 00:26:33.640 --> 00:26:37.959 and made the same rigging system that we had on the train 472 00:26:37.959 --> 00:26:44.040 and attached it virtually to film that scene 473 00:26:44.040 --> 00:26:46.440 Through this, we could produce a scene 474 00:26:46.440 --> 00:26:51.760 where it seamlessly moved from a 2D source to a tunnel source 475 00:26:51.760 --> 00:26:55.399 What you see on the left is an actual tunnel that we filmed 476 00:26:55.399 --> 00:26:59.279 and one on the left is a tunnel video made in 3D 477 00:26:59.279 --> 00:27:03.160 Real tunnels were too dark to show anyting at all 478 00:27:03.160 --> 00:27:05.440 but the virtual tunnel had some lighting 479 00:27:05.440 --> 00:27:09.040 which created the feel of the interior of a tunnel 480 00:27:09.040 --> 00:27:12.679 and we could adjust the placement or brightness of the lighting as well 481 00:27:12.679 --> 00:27:15.519 so we prepared the 482 00:27:15.519 --> 00:27:18.799 tunnel sources in 3D 483 00:27:18.799 --> 00:27:22.160 Another problem that occurred when preparing the source was 484 00:27:22.160 --> 00:27:26.279 we usually filmed driving comps on roads 485 00:27:26.279 --> 00:27:29.200 and when the weather is sunny 486 00:27:29.200 --> 00:27:31.399 and at nights, when the sky was clear 487 00:27:31.399 --> 00:27:33.239 But this time, there were a lot of unpaved roads 488 00:27:33.239 --> 00:27:36.279 and they also wanted the weather to be gloomy as well 489 00:27:36.279 --> 00:27:39.040 so it was very windy 490 00:27:39.040 --> 00:27:41.799 That caused two problems 491 00:27:41.799 --> 00:27:43.440 Since it was windy 492 00:27:43.440 --> 00:27:48.959 the fan in the cameras would stop 493 00:27:48.959 --> 00:27:50.839 and the camera would get heated 494 00:27:50.839 --> 00:27:54.720 causing it to shut down 495 00:27:54.720 --> 00:27:56.799 That made us stop the shooting 496 00:27:56.799 --> 00:28:01.559 and turning it back on extended the source capturing time 497 00:28:01.559 --> 00:28:03.640 So after shooting like this one day 498 00:28:03.640 --> 00:28:07.000 we made a wind shield 499 00:28:07.000 --> 00:28:09.720 so that no wind gets in the way 500 00:28:09.720 --> 00:28:13.000 Of course it wouldn't be the best at keeping the camera cool 501 00:28:13.000 --> 00:28:15.920 but keeping the fan going 502 00:28:15.920 --> 00:28:18.959 was important for keeping the camera cool 503 00:28:18.959 --> 00:28:22.679 so we made this wind shield and used it 504 00:28:22.679 --> 00:28:25.920 And now we have this rig format 505 00:28:25.920 --> 00:28:29.359 that we can attach wind shields on it 506 00:28:29.359 --> 00:28:32.880 This as well, if we hadn't done this shoot 507 00:28:32.880 --> 00:28:36.159 we wouldn't have known about this 508 00:28:36.159 --> 00:28:40.239 but trying capturing sources from poor conditions 509 00:28:40.239 --> 00:28:42.679 helped us discover this issue 510 00:28:42.679 --> 00:28:45.119 Then we finally started the rehearsal 511 00:28:45.119 --> 00:28:47.440 The most important thing during the rehearsal 512 00:28:47.440 --> 00:28:50.039 was the interaction between the two cars 513 00:28:50.039 --> 00:28:54.280 We had background footage we filmed with one car 514 00:28:54.280 --> 00:28:58.000 but having two cars arrayed 515 00:28:58.000 --> 00:29:00.840 was something we couldn't really guess how it would look 516 00:29:00.840 --> 00:29:04.559 And would we be able to keep the speed the same while actually driving 517 00:29:04.559 --> 00:29:08.559 Would we be able to drive them to look like one is racing 518 00:29:08.559 --> 00:29:11.340 and the one behind is trying to overtake 519 00:29:11.340 --> 00:29:12.960 These things were tested 520 00:29:12.960 --> 00:29:15.039 Would it be better to push the car 521 00:29:15.039 --> 00:29:18.320 or would it be better to put it on a motion base 522 00:29:18.320 --> 00:29:20.479 Should we push the first car to the back 523 00:29:20.479 --> 00:29:22.359 or should we push the second car to the front 524 00:29:22.359 --> 00:29:25.000 All these things were thought about 525 00:29:25.000 --> 00:29:29.320 and checking the actual backgrounds and sources 526 00:29:29.320 --> 00:29:31.400 we looked for the right timing 527 00:29:31.400 --> 00:29:34.840 Another thing the director wanted was a foggy area 528 00:29:34.840 --> 00:29:37.840 One of the reasons why we shot on a dark day was 529 00:29:37.840 --> 00:29:41.960 we thought that the day must be cloudy to have a foggy looking lighting 530 00:29:41.960 --> 00:29:46.559 So we filmed a lot of the sources before and after it had rained 531 00:29:46.559 --> 00:29:49.320 and in the site we sprayed fog 532 00:29:49.320 --> 00:29:50.570 so that when we filmed it 533 00:29:50.570 --> 00:29:55.679 with a camera, it would look like it was foggy 534 00:29:55.679 --> 00:29:58.400 Now that the shoot is done 535 00:29:58.400 --> 00:30:00.520 I think the fog helped a lot 536 00:30:00.520 --> 00:30:04.440 because LED wall filming is no matter 3D or 2D 537 00:30:04.440 --> 00:30:06.400 it is projecting something on a flat surface 538 00:30:06.400 --> 00:30:11.400 So it looks more realistic when there are more layers in front of it 539 00:30:11.400 --> 00:30:14.919 Having a lot of layer means having a lot of real objects 540 00:30:14.919 --> 00:30:17.359 and when the camera captures objects on 541 00:30:17.359 --> 00:30:20.039 the LED wall after real objects 542 00:30:20.039 --> 00:30:22.880 the spatial perception from the layers 543 00:30:22.880 --> 00:30:25.880 also create spatial perception for the background as well 544 00:30:25.880 --> 00:30:28.599 D.P. season 2 had only two cars 545 00:30:28.599 --> 00:30:32.440 but the fog created another layer 546 00:30:32.440 --> 00:30:36.039 which created a more realistic spatial perception 547 00:30:36.039 --> 00:30:39.479 We rehearsed the train scene as well 548 00:30:39.479 --> 00:30:42.840 That scene had backgrounds 549 00:30:42.840 --> 00:30:45.119 but the speed of the tunnel scene, 550 00:30:45.119 --> 00:30:47.799 the speed of the lighting the height of it 551 00:30:47.799 --> 00:30:51.640 These were the focus of the tests 552 00:30:51.640 --> 00:30:56.159 and through the rehearsal we settled on where the lights should be 553 00:30:56.159 --> 00:30:57.799 how bright the lights must be 554 00:30:57.799 --> 00:31:00.399 and how far each lights must be 555 00:31:00.399 --> 00:31:02.760 from each other 556 00:31:02.760 --> 00:31:05.280 We brought this back to the unreal engine 557 00:31:05.280 --> 00:31:09.719 and made the final source by re-rendering it 558 00:31:09.719 --> 00:31:13.760 We also did a lot of testing on reflection 559 00:31:13.760 --> 00:31:16.640 During Black Knight as well 560 00:31:16.640 --> 00:31:19.280 while shooting Black Knight that we talked about in the third lecture 561 00:31:19.280 --> 00:31:22.880 there were some issues on the reflection lacking 562 00:31:22.880 --> 00:31:26.559 so this time we tried to find the angle that shows the most reflection 563 00:31:26.559 --> 00:31:28.599 and searched for the right location for the car 564 00:31:28.599 --> 00:31:30.999 Instead, we focused on finding lighting values 565 00:31:30.999 --> 00:31:33.400 That could compromise reflection and moiré 566 00:31:33.400 --> 00:31:35.100 Moiré is 567 00:31:35.100 --> 00:31:37.400 the unique pattern of LED showing 568 00:31:37.400 --> 00:31:40.520 LED is devices lighting up 569 00:31:40.520 --> 00:31:42.370 so when those devices are captured by camera 570 00:31:42.370 --> 00:31:44.760 it sometimes makes wavy patterns 571 00:31:44.760 --> 00:31:48.239 and this is happens more often in reflections 572 00:31:48.239 --> 00:31:51.760 So we tried to minimize moiré in the reflection 573 00:31:51.760 --> 00:31:54.280 and searched for the widest reflection area 574 00:31:54.280 --> 00:31:58.359 which was a challenge that we took during the D.P. season 2 shoot 575 00:31:58.359 --> 00:32:01.359 One thing we earned from this was 576 00:32:01.359 --> 00:32:03.909 LED stages are 577 00:32:03.909 --> 00:32:06.359 being made and used in many occasions 578 00:32:06.359 --> 00:32:11.599 but the LED angle, camera shutter opening angle 579 00:32:11.599 --> 00:32:16.809 that we gained from this shoot can be used 580 00:32:16.809 --> 00:32:19.320 as references no matter what the LED brand or spec is 581 00:32:19.320 --> 00:32:21.359 so I think this was a good experience 582 00:32:21.359 --> 00:32:24.799 Rehearsal for the train was also done once the train was in the set 583 00:32:24.799 --> 00:32:29.599 The speed and color of the sources were the main things 584 00:32:29.599 --> 00:32:31.400 but there was just one problem 585 00:32:31.400 --> 00:32:34.000 You will see later but 586 00:32:34.000 --> 00:32:38.840 right after we enter Daejeon station there was a shade 587 00:32:38.840 --> 00:32:42.679 So we had set the camera for outdoors 588 00:32:42.679 --> 00:32:45.429 and the moment the shade was there 589 00:32:45.429 --> 00:32:48.479 the screen would turn extremely bright or dark 590 00:32:48.479 --> 00:32:50.359 because the iris diaphragm wasn't right 591 00:32:50.359 --> 00:32:52.459 We didn't have auto iris diaphragm 592 00:32:52.459 --> 00:32:53.640 so it was inevitable 593 00:32:53.640 --> 00:32:56.090 but the director wanted the lighting 594 00:32:56.090 --> 00:32:59.440 to be the same even when the shade appeared 595 00:32:59.440 --> 00:33:01.200 after the rehearsal 596 00:33:01.200 --> 00:33:06.039 But we didn't have the time to film it again 597 00:33:06.039 --> 00:33:11.239 so we had to color edit that part 598 00:33:11.239 --> 00:33:13.989 to brighten it a little bit 599 00:33:13.989 --> 00:33:16.880 and make it less awkward when entering Daejeon station 600 00:33:16.880 --> 00:33:19.640 Looking at the final product it isn't too noticeable 601 00:33:19.640 --> 00:33:22.760 but the colors do seem 602 00:33:22.760 --> 00:33:24.960 a little bit burnt off 603 00:33:24.960 --> 00:33:26.880 because there was too less light 604 00:33:26.880 --> 00:33:30.280 That was another issue that we discovered during the rehearsal 605 00:33:30.280 --> 00:33:32.780 I assumed that if we have another situation 606 00:33:32.780 --> 00:33:35.359 where we enter an environment that suddenly gets dark 607 00:33:35.359 --> 00:33:37.559 we would use two sets of cameras 608 00:33:37.559 --> 00:33:40.280 We would set one set to the outdoors 609 00:33:40.280 --> 00:33:43.840 and the other to the darker inside 610 00:33:43.840 --> 00:33:45.990 to get both sources 611 00:33:45.990 --> 00:33:48.039 will be what we'll do 612 00:33:48.039 --> 00:33:51.119 Then we get into the production 613 00:33:51.119 --> 00:33:54.760 This project was unique because 614 00:33:54.760 --> 00:33:56.660 the team doing the post-production 615 00:33:56.660 --> 00:33:59.799 and the virtual production were different teams 616 00:33:59.810 --> 00:34:02.400 and the most important thing was 617 00:34:02.400 --> 00:34:04.750 the communication between the targeted direction of the post-production 618 00:34:04.750 --> 00:34:06.119 and how we should do our part 619 00:34:06.119 --> 00:34:09.769 So we established a way to 620 00:34:09.769 --> 00:34:11.719 periodically communicate with the VFX supervisor 621 00:34:11.719 --> 00:34:16.559 and through this I could see that working together with 622 00:34:16.559 --> 00:34:18.719 another company is possible in a project 623 00:34:18.719 --> 00:34:22.239 You're seeing the train scene shooting 624 00:34:22.239 --> 00:34:23.880 Many different angles 625 00:34:23.880 --> 00:34:26.000 The advantage of the train scene was 626 00:34:26.000 --> 00:34:30.520 that we could film both directions of the actors 627 00:34:30.520 --> 00:34:33.000 and the tunnel effect was also 628 00:34:33.000 --> 00:34:37.520 implemented without an issue 629 00:34:37.520 --> 00:34:40.640 Unlike cars 630 00:34:40.640 --> 00:34:44.719 there were no close ups on the windows 631 00:34:44.719 --> 00:34:47.080 so other artifacts shown on the LED 632 00:34:47.080 --> 00:34:50.719 like light smudging or some moiré forming 633 00:34:50.719 --> 00:34:52.529 wasn't noticeable at all 634 00:34:52.529 --> 00:34:57.679 I think we could focus on the scene itself for the train scenes 635 00:34:57.679 --> 00:34:59.719 For car scenes, the problem of how to move 636 00:34:59.719 --> 00:35:02.640 the two cars that I mentioned earlier 637 00:35:02.640 --> 00:35:06.320 also didn't have much issues since the sprayed fog 638 00:35:06.320 --> 00:35:09.039 created another layer and made the shot 639 00:35:09.039 --> 00:35:12.039 look much more natural 640 00:35:12.039 --> 00:35:15.159 What you're seeing is the second two car scene 641 00:35:15.159 --> 00:35:16.599 where you can see the sports car pass by 642 00:35:16.599 --> 00:35:20.039 That scene looks interesting even now 643 00:35:20.039 --> 00:35:23.000 because the speed of the sports car and the background don't match at all 644 00:35:23.000 --> 00:35:27.000 That scene we didn't settle on what sources to use in the rehearsal 645 00:35:27.000 --> 00:35:30.919 and chose only at the last moment 646 00:35:30.919 --> 00:35:33.400 so the source's frame rate and the 647 00:35:33.400 --> 00:35:35.880 speed wasn't determined at all 648 00:35:35.880 --> 00:35:38.080 which made it a bit awkward 649 00:35:38.080 --> 00:35:42.919 I think that it was a case that 650 00:35:42.919 --> 00:35:44.969 let us know how to set the sources 651 00:35:44.969 --> 00:35:46.880 for cars going in completely 652 00:35:46.880 --> 00:35:49.640 different speeds 653 00:35:49.640 --> 00:35:53.799 This was a summary of D.P. season 2 654 00:35:53.799 --> 00:35:58.200 which is a successful ICVFX case 655 00:35:58.200 --> 00:36:00.119 All shots were given OKs 656 00:36:00.119 --> 00:36:04.880 and all shots were used in the final without editing afterwards 657 00:36:04.880 --> 00:36:08.000 I think this was a project that proved that 658 00:36:08.000 --> 00:36:11.479 for projects like D.P. season 2 with 659 00:36:11.479 --> 00:36:15.840 large scale trains or driving scenes 660 00:36:15.840 --> 00:36:18.719 not only simple dialog but also action or car chasing scenes 661 00:36:18.719 --> 00:36:20.960 can be done in 2D driving comp 662 00:36:20.960 --> 00:36:23.440 I think it is a technology 663 00:36:23.440 --> 00:36:27.559 that can be used in many other contents with the lessons we took from this 664 00:36:27.559 --> 00:36:30.799 I explained about 2D driving comp in this lecture 665 00:36:30.799 --> 00:36:34.039 2D driving comp is in a way 666 00:36:34.039 --> 00:36:37.320 combining ICVFX and sources from existing filming methods 667 00:36:37.320 --> 00:36:40.440 which shows a new production method 668 00:36:40.440 --> 00:36:42.840 while also being the most applicable method 669 00:36:42.840 --> 00:36:44.340 Like this 670 00:36:44.340 --> 00:36:47.469 I hope you can combine the technology you are learning now, 671 00:36:47.469 --> 00:36:49.880 existing filming methods and the content being created 672 00:36:49.880 --> 00:36:52.840 to produce the highest quality 673 00:36:52.840 --> 00:36:56.559 in the shortest time 674 00:36:56.559 --> 00:36:59.039 Thank you for listening 675 00:36:59.558 --> 00:37:06.258 Workflow of 2D driving comp technology Problems of driving comp Absence of source to use driving comp Sync Camera placement 676 00:37:06.258 --> 00:37:13.958 2D driving comp examples Production steps Storyboard analysis LED volume Tech-viz and testing Source production and preparation using trains Rehearsal Production