WEBVTT 00:00.000 --> 00:03.500 Who would have thought I would be faster than him? 00:04.540 --> 00:05.880 That you are faster than me? 00:06.560 --> 00:08.680 You started earlier also, eh? 00:10.960 --> 00:12.160 You're not talking about that? 00:12.500 --> 00:12.940 No. 00:13.000 --> 00:13.580 Oh, okay. 00:13.581 --> 00:14.440 We started at the same time. 00:14.500 --> 00:15.580 Yeah, of course. 00:16.120 --> 00:18.040 We had some time between projects. 00:18.300 --> 00:20.400 Rick proposed to make some tests for Gris Pencil. 00:20.500 --> 00:22.440 It was being developed at the same time. 00:22.480 --> 00:23.800 They were rewriting all the stuff. 00:23.801 --> 00:28.360 And I was interested in doing some kind of 3D NPR anime. 00:28.740 --> 00:29.740 That's how it came. 00:37.920 --> 00:43.720 I wanted to capture that essence of 90s action heroes and 90s animes. 00:43.920 --> 00:45.760 That's what I grew up watching. 00:45.960 --> 00:48.420 A lot of Dragon Ball, a lot of Cowboy Bebop. 00:48.620 --> 00:51.540 The idea was to have a buff cow punching robots. 00:51.720 --> 00:53.640 Obliko started with some initial sketches. 00:53.940 --> 00:57.520 Based off those sketches, I started mocking up just a simple model. 00:57.521 --> 01:01.420 Me and Vivian got kind of like a back and forth, where she would do some sculpting, 01:01.500 --> 01:04.020 some adjustments, and she made it a lot more organic looking. 01:04.420 --> 01:07.400 I mean, it's very, very easy to work for Obliko. 01:07.600 --> 01:10.340 He pretty much says, I want that. 01:10.580 --> 01:11.300 This is good. 01:11.540 --> 01:12.540 This is bad. 01:12.580 --> 01:13.360 Make it more buff. 01:13.540 --> 01:16.640 One of the important things was, of course, Cowboy is super buff. 01:16.820 --> 01:20.460 So the musculature needed to look good and believable. 01:20.660 --> 01:23.906 So we spent some time on looking at anatomy and 01:23.907 --> 01:26.220 trying to get that buffness shining through. 01:26.221 --> 01:31.580 Also, in the topology itself, some of the line work needed to be correctly placed in 01:31.581 --> 01:36.640 the UVs, meaning that we had to set up to add extra lines that matched the curvature 01:36.641 --> 01:38.620 of the bicep and the triceps, stuff like that. 01:38.660 --> 01:41.920 And the face, especially the brow area, had some 01:41.921 --> 01:44.480 iterations where it needed to be a lot more flat. 01:44.620 --> 01:47.760 Then we had some outlines with Crease Pencil 3 line art. 01:47.860 --> 01:51.940 Also some just marked lines in the mesh where Obliko just went into the character 01:51.941 --> 01:56.540 and just said, OK, I want this line here to be always showing up in the render so 01:56.541 --> 01:59.040 it pronounces some muscle on the arm or something. 01:59.280 --> 02:03.440 So with the robot, Francesco challenged us to do it in as little time as possible, 02:03.520 --> 02:07.620 which he imagined that to be one day, two days, something like that. 02:07.660 --> 02:10.840 And I had only done really, really quick scribbles. 02:11.120 --> 02:13.179 We needed something more proportionally 02:13.180 --> 02:15.540 extreme, something that fits more with Cowboy. 02:15.680 --> 02:19.168 So I started setting up a character, something similar in 02:19.169 --> 02:21.980 size to what we were already using for the layout animation. 02:21.981 --> 02:24.560 We had the charge robot before that. 02:24.600 --> 02:26.200 That was basically just a stand-in. 02:26.300 --> 02:31.100 So he set up this very simple base, and then I started sculpting over it 02:31.101 --> 02:35.220 because I at least could make sense of my horrible quick scribbles. 02:35.280 --> 02:38.611 We also looked at action figurines a lot because 02:38.612 --> 02:41.581 they have to move without the parts deforming. 02:41.620 --> 02:45.620 At first, it seemed crazy to do it in just two days, but then we managed quite well. 02:45.860 --> 02:50.040 So the urban environment was actually based off Snow Procure short. 02:50.041 --> 02:55.180 I basically simplified the shading and I placed some buildings, and then that's it. 02:55.220 --> 02:59.200 We just created a simple background render and Vivian painted over it. 02:59.340 --> 03:02.580 And that was just like a 2D panorama that was being animated. 03:02.920 --> 03:06.180 Bo and Pabliko pretty much prepared the files for me. 03:06.260 --> 03:09.680 So I just got these renders that I could just paint over. 03:10.180 --> 03:11.980 I mean, they had one or two layers. 03:12.020 --> 03:13.060 It was very simple. 03:13.420 --> 03:18.280 And for the ground, I created a simple setup where Pabliko could skew the rooftop 03:18.281 --> 03:24.060 to emphasise or make the background seem more extreme, like a forced perspective. 03:24.820 --> 03:27.620 I spent all my brain cells doing this. 03:28.340 --> 03:29.980 Are you ready to be done? 03:30.600 --> 03:32.020 I'm ready to be... 03:32.860 --> 03:34.200 Are you ready to be done? 03:34.300 --> 03:35.520 Yes, I am. 03:35.600 --> 03:35.920 Almost. 03:36.320 --> 03:39.640 Overall, we kept the shading for Cowboy pretty simple. 03:40.020 --> 03:43.264 Basically just have a two-tone shading of just light 03:43.265 --> 03:45.520 and dark for different materials of the character. 03:45.521 --> 03:48.873 And it's really just very simple two-tone shading 03:48.874 --> 03:52.160 based on the Shader to RGB pipeline in Eevee. 03:52.220 --> 03:56.460 Quite different to what we did for Wing -It, where the idea was to try and come up 03:56.461 --> 03:59.260 with a smart system that just works by itself. 03:59.440 --> 04:01.280 So we wouldn't have to do too much manual work. 04:01.360 --> 04:04.420 But with Cowboy, we kind of went the opposite way of saying, OK, we want to 04:04.421 --> 04:08.160 just have a simple base setup, but then have a lot of manual control over it. 04:08.200 --> 04:11.159 Just set up some parameters so we could change the shadow 04:11.160 --> 04:13.600 threshold on the fly also for the different objects. 04:13.601 --> 04:18.320 The lighting itself, like the baseline was super simple, just basically a sunlight to 04:18.321 --> 04:21.309 give the main lighting direction for the main 04:21.310 --> 04:23.800 shadow paths that then would be cleaned up manually. 04:24.080 --> 04:25.720 We use Grease Pencil in different parts. 04:26.060 --> 04:27.060 All the effects are 2D. 04:27.100 --> 04:30.260 They are complex in the drawing, but they are not complex on technique 04:30.261 --> 04:34.480 because they were like basically just a plane between the characters and the cameras. 04:34.600 --> 04:38.140 The most challenging thing that would push more further was the sticky GP. 04:38.260 --> 04:41.980 The idea is essentially that with Grease Pencil 3, we have the ability to use 04:41.981 --> 04:45.520 geometry nodes, and with that we have a lot more ability to make our own tooling. 04:45.720 --> 04:51.000 We created a tool that we were able to draw over 3D mesh with the strokes and 04:51.001 --> 04:53.000 they will attach to the mesh and they will follow. 04:53.180 --> 04:58.460 For example, to stick the Grease Pencil data that is drawn in on the frame to the 04:58.461 --> 05:03.140 underlying surface so that it just follows that surface and that deformation of the 05:03.141 --> 05:07.360 character throughout the rest of the shot, which is really useful to have fine 05:07.361 --> 05:11.460 control over the shadow lines because a lot of the time you want them to kind of 05:11.461 --> 05:14.600 be consistent to resemble traditional 2D look. 05:14.780 --> 05:20.520 We were able to draw over the model and not have to draw each frame with this. 05:20.660 --> 05:23.767 You could draw once and then if the poses were 05:23.768 --> 05:26.821 similar, you could hold that for many, many frames. 05:26.960 --> 05:29.330 It was mostly for shadows and tried to make 05:29.331 --> 05:31.560 everything more blocky and more like hand -drawn. 05:31.840 --> 05:35.380 Pablico went over everything to clean up the shadow lines exactly to get the kind 05:35.381 --> 05:39.620 of look that he wanted, which was a lot more angular than the original gave us. 05:39.621 --> 05:40.500 It was a lot more smooth. 05:40.501 --> 05:45.620 Wherever it was a mismatch, he could just go in and then redraw it and stick it for 05:45.621 --> 05:49.580 the rest of the shot to see how long it sticks well and then at some point when 05:49.581 --> 05:51.810 the arm would rotate too much, you need to 05:51.811 --> 05:53.800 redraw it and then stick it again and everything. 05:54.060 --> 05:56.420 The workflow for him was pretty streamlined in that sense. 05:56.580 --> 05:58.120 Your color void has been done. 05:58.700 --> 05:59.620 Can we see the file? 05:59.720 --> 06:01.120 Who said that it was a competition? 06:02.080 --> 06:03.080 Never hear anything. 06:03.740 --> 06:05.240 You're the competition guy. 06:05.560 --> 06:06.920 Yeah, but not in this thing. 06:07.000 --> 06:08.000 Pablico! 06:08.700 --> 06:10.801 You're not even... Ah! 06:12.480 --> 06:13.760 It's not a competition. 06:13.820 --> 06:15.400 I'm just playing, it's just a sleeve. 06:15.780 --> 06:16.520 Yeah, yeah. 06:16.660 --> 06:22.120 A while ago, Matt and me together worked on an add-on that gives the animators the 06:22.121 --> 06:27.020 ability to create basically shape keys on the fly where they could just press a 06:27.021 --> 06:32.600 button and then the already posed and animated character would be sculptable. 06:32.780 --> 06:34.966 That they could just go into sculpt mode and then 06:34.967 --> 06:37.780 change the shape with the familiar sculpting tools. 06:37.781 --> 06:42.740 Then for that frame or however long they want that to be an influence they could 06:42.741 --> 06:46.540 apply that shape key to get the exact shape that they want with sculpting tools. 06:46.800 --> 06:49.820 Because the rigging of the character was not really great. 06:49.920 --> 06:51.060 We did it in a rush. 06:51.200 --> 06:54.007 By pushing with the shape keys I could just 06:54.008 --> 06:56.220 make the pose I wanted by just pushing the mesh. 06:56.360 --> 07:01.160 The big benefit of that is that you don't need to rely on a very complex rig that 07:01.161 --> 07:05.160 can hit all of the different shapes that you might possibly want to hit. 07:05.161 --> 07:10.100 So you don't really rely on a very involved rigging process and iteration 07:10.101 --> 07:13.440 cycle between the animator and the rigger to really make sure they can hit all of 07:13.441 --> 07:17.540 those extreme poses that might be very unique for that one shot but rather you 07:17.541 --> 07:19.384 can just sculpt it into shape and it just works 07:19.385 --> 07:21.160 and you don't have to touch the rig at all. 07:21.460 --> 07:24.555 We encountered some issues with it which we took some 07:24.556 --> 07:26.720 notes of things we want to improve in the future. 07:26.820 --> 07:31.180 Mainly that was the performance because Publico really went in adding lots and 07:31.181 --> 07:35.780 lots of shape keys on the fly because the rig wasn't doing it for some of the poses. 07:35.980 --> 07:38.427 And then at some point the performance started dragging 07:38.428 --> 07:40.500 so it wasn't really visible in real time anymore. 07:40.700 --> 07:42.760 But besides that it was pretty smooth sailing. 07:42.940 --> 07:44.819 I mean we were working with the alpha version of 07:44.820 --> 07:47.460 Blender so there's always crashes and bugs happening. 07:51.570 --> 07:52.330 Press anything. 07:52.590 --> 07:53.970 I just f***ing click it. 07:54.270 --> 07:55.830 Well the timing over there was perfect. 07:56.090 --> 07:57.710 They're very understandable of course. 07:57.850 --> 08:02.110 They know that things can't get fixed from one minute to the other. 08:02.150 --> 08:06.432 They're often actually not that pushy enough where they deal 08:06.433 --> 08:10.070 with problems but they don't tell us developers about them. 08:10.210 --> 08:13.550 There are some cases where they don't really know if it's supposed to work that 08:13.551 --> 08:17.830 way or not and then they just work around the issue like Publico does a lot of times 08:17.831 --> 08:19.584 it just brute forces the problem without telling 08:19.585 --> 08:22.171 us even though it might be a very easy fix. 08:28.870 --> 08:35.050 I mean I think this sticky GP is kind of a product of brute forcing because instead 08:35.051 --> 08:40.210 of going to somebody and touch the normals or whatever for the shadows and everything 08:40.211 --> 08:43.930 I was like well if I can do it by myself then I don't have to ask anybody. 08:44.330 --> 08:48.050 I kind of like to just do it and later ask. 08:48.170 --> 08:53.210 For both of those tools it basically means that any production where you want to put 08:53.211 --> 08:59.610 more of a focus on the clean up the manual involved process of people actually going 08:59.611 --> 09:04.190 over the shots going over the frames and pushing them to the best they can be 09:04.191 --> 09:08.290 rather than investing a lot into pre -production and getting all of the little 09:08.291 --> 09:12.090 features that you might want to possibly have later on already in there and then 09:12.091 --> 09:13.990 have everything run smoothly automatically. 09:14.330 --> 09:17.590 Both of those tools are a pretty good way to go so that's kind of what we were 09:17.591 --> 09:20.810 trying to push with this little project and that worked pretty well. 09:21.010 --> 09:23.185 I think that's basically the biggest thing we 09:23.186 --> 09:25.410 are going to use in the future for all the NPR. 09:25.630 --> 09:29.050 I think we are in a good path but we will see. 09:31.570 --> 09:32.870 Oh, so professional, eh?