1 00:00:25.673 --> 00:00:32.596 I will talk about animation AI application cases 2 00:00:33.240 --> 00:00:35.739 The fact that AI is eternally evolving the way 3 00:00:36.857 --> 00:00:41.365 how stories are delivered is a well know fact 4 00:00:41.959 --> 00:00:46.521 AI prompt literally changes how our brain processes information 5 00:00:46.521 --> 00:00:49.104 and how it views the world 6 00:00:49.430 --> 00:00:54.119 In this lecture, I will tell you how AI is changing creative storytelling 7 00:00:54.119 --> 00:00:59.306 and how it is in real life right now 8 00:01:00.129 --> 00:01:03.980 Storytelling using AI 9 00:01:04.555 --> 00:01:10.546 The fact that AI will change how story is delivered enternally is not a secret 10 00:01:11.070 --> 00:01:15.748 How exactly is AI changing film production? 11 00:01:16.629 --> 00:01:19.650 AI helps all people become 12 00:01:19.650 --> 00:01:22.305 curators of artworks 13 00:01:22.899 --> 00:01:27.535 Before, everything had to be made from the beginning, all new 14 00:01:28.040 --> 00:01:31.213 So not only the ability to create a work 15 00:01:31.807 --> 00:01:34.505 but also the taste has so important 16 00:01:35.040 --> 00:01:40.969 And now, everyone’s creativity has increased at least more than 50 times 17 00:01:41.959 --> 00:01:45.212 Now is an era where even the people couldn’t produce movies 18 00:01:45.212 --> 00:01:48.338 now can 19 00:01:49.160 --> 00:01:54.628 And those who were film producers before can now use other methods 20 00:01:56.559 --> 00:01:59.311 to create stories that were impossible before 21 00:02:01.519 --> 00:02:05.006 Without relying on Hollywood greenlight 22 00:02:06.343 --> 00:02:08.910 they can actually produce a project 23 00:02:09.910 --> 00:02:14.566 I think that thanks to AI tools, more creative stories can be born 24 00:02:15.289 --> 00:02:18.744 We can gather so much more information than any other time before 25 00:02:19.249 --> 00:02:22.799 and I think that this isn’t something to fear for 26 00:02:23.393 --> 00:02:26.834 but something that will help everyone and benefit us 27 00:02:28.596 --> 00:02:34.536 So the most important question, will AI be able to replace your workplace? 28 00:02:35.625 --> 00:02:39.160 I would say ‘no’ 29 00:02:40.160 --> 00:02:42.854 It is not AI replacing your work 30 00:02:43.240 --> 00:02:48.797 but rather more of the possibility of your work being asked to be done in this way increasing 31 00:02:49.519 --> 00:02:53.285 These days, most jobs require computers 32 00:02:53.800 --> 00:02:59.811 Not long from now, AI will be everywhere in our everyday work flow 33 00:03:00.039 --> 00:03:03.275 and it won’t be something as new or scary as it is now 34 00:03:03.800 --> 00:03:05.657 Then a bigger question pops up 35 00:03:06.211 --> 00:03:08.758 Will AI destroy Hollywood? 36 00:03:10.160 --> 00:03:13.241 I’d say ‘no’ as well 37 00:03:15.340 --> 00:03:20.488 But I do think that the entering barrier is lower than any other time 38 00:03:21.349 --> 00:03:26.285 Things that couldn’t be said before is now able to be said by anyone 39 00:03:27.840 --> 00:03:30.781 For example, when thinking about the phase of industrial revolution 40 00:03:30.781 --> 00:03:32.520 I’ll take the example of sweaters 41 00:03:32.520 --> 00:03:36.867 Let’s say we make 10 sweaters a day 42 00:03:37.709 --> 00:03:42.759 Someone could make one sweater a day by knitting it 43 00:03:42.759 --> 00:03:47.740 but after the industrial revolution, now, we can produce 5,000 sweaters a day 44 00:03:48.156 --> 00:03:50.674 with the technology that we have 45 00:03:51.199 --> 00:03:56.051 I think the creativity using AI is the same 46 00:03:57.160 --> 00:04:00.839 A thought by one person is through AI 47 00:04:00.839 --> 00:04:03.722 able to make hundreds of thousands of diversity 48 00:04:04.821 --> 00:04:08.956 AI basically makes everyone a super hero 49 00:04:09.431 --> 00:04:11.950 and I think there is no reason to fear this 50 00:04:13.960 --> 00:04:19.327 Let’s take a look at some cases where it shows how AI can 51 00:04:20.436 --> 00:04:22.787 actually benefit people’s creativity 52 00:04:23.500 --> 00:04:27.733 Through AI, an artist can visualized 53 00:04:28.109 --> 00:04:29.828 what their final work will look like 54 00:04:30.760 --> 00:04:36.152 This technology can analyze and help the artist 55 00:04:36.330 --> 00:04:38.091 even if they are not stopped by the wall of creativity 56 00:04:39.200 --> 00:04:41.702 They can get feedback 57 00:04:41.880 --> 00:04:44.519 without waiting for others to read the script 58 00:04:45.440 --> 00:04:50.078 It can scan the creativity of the text and its originality 59 00:04:50.850 --> 00:04:55.311 and can be used to check how true it is to the original work 60 00:04:56.519 --> 00:05:01.604 It can also spot continuity error that’s hard to spot 61 00:05:02.000 --> 00:05:07.695 It can also perform AI research scanning and summarizing millions of documents 62 00:05:08.150 --> 00:05:10.629 doing the jobs of the following 63 00:05:11.480 --> 00:05:17.617 Introverted people can also introduce a project visually 64 00:05:18.270 --> 00:05:21.445 They do not have to become a extroverted, cool pitch machine 65 00:05:21.920 --> 00:05:23.920 To editors, this tool 66 00:05:24.158 --> 00:05:27.395 has the power to change their work immediately 67 00:05:27.840 --> 00:05:32.546 A real life movie project can be made without actual hands on work 68 00:05:33.160 --> 00:05:36.661 Camera color grading can easily 69 00:05:36.661 --> 00:05:39.151 produce consistent colors 70 00:05:40.399 --> 00:05:43.683 We can remove and replace objects from the background 71 00:05:44.000 --> 00:05:47.679 According to the prompt, desired track can be 72 00:05:47.679 --> 00:05:49.516 produced with one’s exact preferences 73 00:05:50.239 --> 00:05:55.238 Also, by cloning actors’ voice and inputting desired lines 74 00:05:55.505 --> 00:05:57.178 audio can be produced 75 00:05:58.079 --> 00:06:00.723 With just one click, the background can be switched 76 00:06:02.079 --> 00:06:05.369 Video resolution can be improved, increasing the quality 77 00:06:06.300 --> 00:06:09.771 Frame speed can be changed post production 78 00:06:10.880 --> 00:06:15.485 AI can be used in various tools from storyboard making 79 00:06:16.000 --> 00:06:18.873 to complex processes 80 00:06:20.200 --> 00:06:25.536 Next I will tell you the fields I expect to use AI a lot 81 00:06:26.279 --> 00:06:33.640 I think that AI will grow in fields where 82 00:06:33.640 --> 00:06:39.740 computer or mechanic analysis is possible 83 00:06:39.740 --> 00:06:42.731 and where data gathering is convenient 84 00:06:42.860 --> 00:06:45.947 Idea generation and script writing 85 00:06:46.640 --> 00:06:52.824 Scene blocking, continuity error spotting, casting proposals 86 00:06:54.012 --> 00:06:58.829 Realistic storyboard production, converting images into moving environments 87 00:06:59.681 --> 00:07:05.502 Setting drone filming simulations, detail shots and more 88 00:07:06.275 --> 00:07:08.648 The fields that were 89 00:07:09.421 --> 00:07:15.898 done mechanically by programmers or 90 00:07:15.898 --> 00:07:20.182 storyboard artists before 91 00:07:20.559 --> 00:07:26.655 will be made possible to do 92 00:07:27.170 --> 00:07:32.687 with AI, automatically 93 00:07:33.489 --> 00:07:37.534 Among them, the most important ones in videography 94 00:07:37.870 --> 00:07:39.901 will be removing backgrounds 95 00:07:39.901 --> 00:07:46.208 or producing animation of characters using data 96 00:07:48.020 --> 00:07:52.040 and drawing storyboards 97 00:07:52.040 --> 00:07:58.016 I think these field of animation or video production 98 00:07:58.600 --> 00:08:01.778 will use AI a lot 99 00:08:02.986 --> 00:08:07.204 But as I told you before 100 00:08:08.768 --> 00:08:17.935 this will help human 101 00:08:18.529 --> 00:08:21.919 by minimizing effort of minimal people 102 00:08:24.216 --> 00:08:28.338 without taking or intruding human jobs 103 00:08:29.417 --> 00:08:32.682 Just like the industrial revolution, before AI 104 00:08:34.791 --> 00:08:39.992 a human could make one animation work 105 00:08:40.705 --> 00:08:43.331 in 1 or 2 years 106 00:08:43.331 --> 00:08:47.889 while now, it is seen that tens and hundreds 107 00:08:47.889 --> 00:08:50.862 of animations can be made 108 00:08:52.040 --> 00:08:55.720 So managing these is very important 109 00:08:55.720 --> 00:09:00.040 so I think that it is very important 110 00:09:01.921 --> 00:09:05.627 for human to manage the AI and continue to play the role of directors 111 00:09:06.568 --> 00:09:14.873 So although simple labor has decreased 112 00:09:15.130 --> 00:09:18.766 I’m expecting roles of making creatuce decisions 113 00:09:18.766 --> 00:09:22.430 will rather increase greatly 114 00:09:23.559 --> 00:09:27.030 What AI has hard time doing include 115 00:09:28.960 --> 00:09:32.683 improving creative scripts, showing empathy, 116 00:09:33.297 --> 00:09:38.960 creating a video from the beginning, maintaining consistency, creating lively movements, 117 00:09:38.960 --> 00:09:42.670 moving human facial expressions, finger rendering 118 00:09:43.482 --> 00:09:48.180 These would be the examples 119 00:09:48.180 --> 00:09:53.011 These are things that humans need to 120 00:09:53.179 --> 00:09:56.069 decide in detail 121 00:09:56.069 --> 00:10:00.820 so even though AI can make it roughly 122 00:10:01.077 --> 00:10:05.606 deciding and choosing this 123 00:10:05.606 --> 00:10:08.662 must be done by humans in my opinion 124 00:10:10.147 --> 00:10:13.959 And considering the technology of the current time 125 00:10:15.504 --> 00:10:21.706 ideally, looking into the fields of visual AI, video, movie projects 126 00:10:23.617 --> 00:10:28.578 visual film pitch, concept trailer, 127 00:10:29.063 --> 00:10:35.441 scene test and previsualization and storyboard, music video, experiment projects, ads, 128 00:10:35.857 --> 00:10:43.436 concert visuals, science fictions, bio organic stories, shot and detail settings, 129 00:10:43.436 --> 00:10:46.551 parody and comedy etc. are there 130 00:10:46.948 --> 00:10:51.681 These, considering the realistic technologies 131 00:10:52.691 --> 00:10:56.772 are the AI content fields that are being used 132 00:10:57.904 --> 00:11:04.852 Realistically, even if they might not be able to complete a project, 133 00:11:05.208 --> 00:11:13.360 it is being used for pitching or trailer or for research 134 00:11:13.360 --> 00:11:16.566 for now 135 00:11:17.377 --> 00:11:21.308 AI tools that can be used in animation production 136 00:11:22.209 --> 00:11:25.451 Next, I will tell you about how these AI tools can be 137 00:11:25.451 --> 00:11:30.416 used in animation production 138 00:11:31.000 --> 00:11:36.353 focusing on pre-production 139 00:11:37.324 --> 00:11:41.123 These AI tools are, as you may know, 140 00:11:42.688 --> 00:11:48.908 aren't something that you can download or purchase the software 141 00:11:49.136 --> 00:11:51.320 but instead you must subscribe in a website 142 00:11:51.320 --> 00:11:55.516 or used is in remote 143 00:11:55.516 --> 00:11:58.354 in the website server 144 00:11:59.225 --> 00:12:02.118 That's how you'll use them 145 00:12:03.375 --> 00:12:07.792 So when explaining AI 146 00:12:08.099 --> 00:12:13.719 I think it will be good for you to see the software websites 147 00:12:13.719 --> 00:12:18.023 that are out now and see what functions there are 148 00:12:18.736 --> 00:12:22.361 and consider which is good and right for you 149 00:12:23.173 --> 00:12:24.397 and use it 150 00:12:25.040 --> 00:12:33.873 First, in Runway Gen2 you can see these softwares 151 00:12:33.873 --> 00:12:36.969 and or subscribe to certain websites 152 00:12:37.415 --> 00:12:41.359 You would not subscribe to all websites 153 00:12:41.359 --> 00:12:47.021 but you could try the free trial first and then 154 00:12:49.080 --> 00:12:54.811 see what would be useful to you in production 155 00:12:56.174 --> 00:13:00.247 There is a reason to not rely on these too much 156 00:13:00.880 --> 00:13:04.799 Not long ago, the release was announced 157 00:13:04.799 --> 00:13:13.176 Sora, an image generation AI tool by OpenAI was introduced 158 00:13:14.257 --> 00:13:16.956 When you look at the results of that 159 00:13:17.233 --> 00:13:24.915 you'll be able to see much better quality results 160 00:13:25.410 --> 00:13:33.414 compared to other AI products from Runway, Midjourney, or Dolly 161 00:13:33.840 --> 00:13:40.219 Though it is not released yet, AI companies are 162 00:13:40.219 --> 00:13:45.796 continuing to make these AI imaging tools 163 00:13:46.420 --> 00:13:49.238 so in addition to tools you are using now 164 00:13:49.238 --> 00:13:51.640 there will be more coming out 165 00:13:51.640 --> 00:13:55.110 So when you're using them 166 00:13:55.229 --> 00:13:59.589 consider that there are also these things 167 00:14:01.659 --> 00:14:05.612 and once there comes out the perfect tool later 168 00:14:05.612 --> 00:14:08.280 we will use that I expect 169 00:14:08.280 --> 00:14:12.374 We will used them like Unreal Engine or Maya of now 170 00:14:12.374 --> 00:14:14.624 but since we're in a transition period right now 171 00:14:15.654 --> 00:14:20.238 you could approach it in a trying out many things sense 172 00:14:21.080 --> 00:14:32.665 So when you go to Runway and sign up 173 00:14:33.229 --> 00:14:40.119 you'll see this screen and if you click Start Generating 174 00:14:40.119 --> 00:14:46.491 you'll be able to make videos with pictures 175 00:14:46.679 --> 00:14:50.919 This Runway is a video creating tool 176 00:14:50.919 --> 00:14:57.415 So if you have pictures or drawings of something 177 00:14:57.910 --> 00:15:03.008 those are just still cuts that don't move 178 00:15:04.088 --> 00:15:08.822 But using Runway we can adjust the camera here 179 00:15:09.941 --> 00:15:14.979 or add simple animation movements 180 00:15:15.771 --> 00:15:17.479 That's how it works 181 00:15:17.479 --> 00:15:23.431 So if I add a photo by dragging it in 182 00:15:25.035 --> 00:15:28.799 and then write how I want this image to move 183 00:15:28.799 --> 00:15:34.512 in the option window, the prompt window down here 184 00:15:36.215 --> 00:15:38.479 you'll be able to make a video 185 00:15:38.480 --> 00:15:41.582 For example I say Cinematic 186 00:15:55.820 --> 00:16:01.617 I put 'Cinematic shot of this picture with panning left' as the prompt 187 00:16:03.161 --> 00:16:05.417 and click Generate 188 00:16:06.130 --> 00:16:10.946 It takes a bit of time generating 189 00:16:12.480 --> 00:16:13.520 so if you subscribe 190 00:16:13.520 --> 00:16:18.058 it will take less time to make the video 191 00:16:18.781 --> 00:16:21.320 So for now 192 00:16:21.320 --> 00:16:24.655 if you use this photo to make this video 193 00:16:25.239 --> 00:16:29.967 well it won't be able to move the photo but 194 00:16:30.730 --> 00:16:38.710 in this platform, it will calculate the depth of the photo 195 00:16:38.799 --> 00:16:41.827 and after calculating the depht 196 00:16:42.441 --> 00:16:45.520 the camera, when I say camera 197 00:16:45.520 --> 00:16:47.621 I mean that moving the photo to left and right 198 00:16:47.760 --> 00:16:49.454 Moving it left and right, 199 00:16:49.889 --> 00:16:55.991 due to the depth the things in the back moves slower 200 00:16:55.991 --> 00:16:59.656 and the things in the front moves faster 201 00:17:00.765 --> 00:17:02.840 Adjusting the speed 202 00:17:02.840 --> 00:17:05.075 so that when we see the result 203 00:17:06.481 --> 00:17:09.645 it seems like it is moving very 3 dimensionally 204 00:17:11.437 --> 00:17:16.550 There is another video added 205 00:17:18.491 --> 00:17:26.839 but you can see me moving in the video made with my photo 206 00:17:26.839 --> 00:17:30.077 I am moving a little 207 00:17:31.720 --> 00:17:34.520 but there aren't too big movements 208 00:17:34.520 --> 00:17:39.760 with usually the water sparkling 209 00:17:39.760 --> 00:17:41.475 or the hair flying 210 00:17:41.901 --> 00:17:44.706 That's what happens most of the time 211 00:17:45.350 --> 00:17:52.051 To create more diverse movements 212 00:17:52.051 --> 00:17:54.621 it will need many kinds of data 213 00:17:54.730 --> 00:17:56.626 about the movements that I have 214 00:17:57.616 --> 00:18:00.154 Later there will be these things 215 00:18:00.471 --> 00:18:05.922 but for now, this is the video that can be made with just this one photo 216 00:18:06.209 --> 00:18:11.267 So when making it, in this option window 217 00:18:11.870 --> 00:18:13.747 you can choose how you'll have the panning 218 00:18:15.499 --> 00:18:20.126 If I say I'll have a lot of panning and regenerate it 219 00:18:20.492 --> 00:18:25.581 Panning means moving the camera from left to right and up and down 220 00:18:25.581 --> 00:18:32.399 We call this camera movement panning 221 00:18:32.399 --> 00:18:37.450 So if I say that I'll have a lot of panning it will move a lot from left to right 222 00:18:38.311 --> 00:18:41.840 So the terms appearing here, 223 00:18:43.048 --> 00:18:47.480 if you get to know about the terms 224 00:18:47.480 --> 00:18:50.452 used in movie cameras 225 00:18:50.719 --> 00:18:52.681 it will be very convenient to use this 226 00:18:53.731 --> 00:18:57.520 I once saw in the news 227 00:18:57.520 --> 00:19:04.167 Jensen Huang of NVIDIA saying 228 00:19:05.870 --> 00:19:07.720 'What you have to do now is 229 00:19:07.720 --> 00:19:10.012 nothing about learning technology 230 00:19:11.210 --> 00:19:13.954 You need to know a lot about humanities' 231 00:19:14.360 --> 00:19:17.180 I heard that 232 00:19:17.180 --> 00:19:21.720 So now, technological things 233 00:19:21.720 --> 00:19:24.087 will all be done by AI or computers 234 00:19:24.651 --> 00:19:29.781 and what humans must do is the prompt 235 00:19:30.762 --> 00:19:33.297 According to how well the prompt is written 236 00:19:34.386 --> 00:19:39.556 the video's quality will be high or low 237 00:19:40.090 --> 00:19:42.236 so in a sense we'll need a lot of 238 00:19:42.661 --> 00:19:46.109 humanities knowledge 239 00:19:47.168 --> 00:19:48.836 Taking a look at this video 240 00:19:49.311 --> 00:19:52.141 as I added more panning 241 00:19:54.260 --> 00:19:59.337 you can see that there are much more movements here 242 00:20:00.961 --> 00:20:03.881 Like this, making these shots one by one 243 00:20:03.881 --> 00:20:07.505 and you extract shots from here 244 00:20:07.960 --> 00:20:12.549 And editing what you extracted 245 00:20:13.251 --> 00:20:15.104 you could make an ad video 246 00:20:15.490 --> 00:20:20.151 or short animations for pitching 247 00:20:20.151 --> 00:20:22.693 or things like that 248 00:20:24.129 --> 00:20:25.884 Next is Pika Labs 249 00:20:26.240 --> 00:20:30.396 When you go to Pika Labs website 250 00:20:30.921 --> 00:20:38.732 you can see these example results 251 00:20:39.821 --> 00:20:43.781 and these are also made using 252 00:20:43.781 --> 00:20:49.098 one picture or a drawing 253 00:20:49.098 --> 00:20:53.939 to make an animation 254 00:20:55.731 --> 00:20:59.305 So putting these images 255 00:21:08.651 --> 00:21:11.599 and you could also add lip syncing 256 00:21:11.599 --> 00:21:22.931 So listening to the voices of the character you want 257 00:21:24.030 --> 00:21:27.933 and write the line you want in the lip sync 258 00:21:41.221 --> 00:21:45.505 and then if you do Generate Voice 259 00:21:51.158 --> 00:21:55.927 it will automatically 260 00:21:57.669 --> 00:22:00.756 make it in that voice, an audio 261 00:22:01.558 --> 00:22:05.837 without a voice actor 262 00:22:06.500 --> 00:22:09.600 So do 'attach and continue' 263 00:22:09.600 --> 00:22:14.486 and if you click that you'll generate 264 00:22:18.050 --> 00:22:19.949 a video is being generated now 265 00:22:21.711 --> 00:22:26.682 All the AI tools released until now are in this format 266 00:22:26.830 --> 00:22:31.545 so it is very easy to access 267 00:22:34.451 --> 00:22:42.495 So you can decide the final resolution of this video 268 00:22:46.901 --> 00:22:49.488 and the video length as well 269 00:22:55.320 --> 00:22:57.705 So now we have a video 270 00:23:00.071 --> 00:23:02.160 It became a very funny video 271 00:23:02.160 --> 00:23:08.447 but since I used a photo where I have my mouth covered 272 00:23:09.090 --> 00:23:11.481 the lip sync is a bit weird 273 00:23:11.481 --> 00:23:15.665 So if you listen to the sound 274 00:23:20.279 --> 00:23:23.803 the video is something like this 275 00:23:25.130 --> 00:23:26.505 Simple lip sync 276 00:23:27.050 --> 00:23:30.625 But as you can see it is a bit awkward 277 00:23:31.101 --> 00:23:34.010 Even if it was a picture 278 00:23:36.990 --> 00:23:39.737 where I did not have my mouth covered 279 00:23:41.964 --> 00:23:46.061 the movements of the muscles when the mouth moves 280 00:23:46.061 --> 00:23:49.056 won't be applied right 281 00:23:49.551 --> 00:23:54.380 so it looks like only the mouth is opening and closing when lip syncing 282 00:23:54.380 --> 00:23:57.721 I don't know about the case of 2D animations, 283 00:23:57.721 --> 00:24:04.403 but it can be very awkward since this is real life photo 284 00:24:04.780 --> 00:24:07.339 So if you want to make lip sync animation 285 00:24:07.339 --> 00:24:11.106 with Pika or other tools 286 00:24:11.680 --> 00:24:15.680 using 2D images will give you 287 00:24:15.680 --> 00:24:19.446 relatively better results 288 00:24:20.139 --> 00:24:26.681 We took a look at how to make videos from photos using 289 00:24:26.681 --> 00:24:29.904 Runway and Pika, with prompts 290 00:24:30.340 --> 00:24:33.059 These are these kinds of tools 291 00:24:34.128 --> 00:24:36.240 Runway is used a lot 292 00:24:36.240 --> 00:24:38.723 Runway has 'Prompt to video' 293 00:24:38.990 --> 00:24:41.215 'Video to stylized video' 294 00:24:42.611 --> 00:24:44.640 and 'Image to video' 295 00:24:45.551 --> 00:24:47.711 and others as well 296 00:24:48.691 --> 00:24:52.982 If you do 'Video to stylized video' 297 00:24:53.210 --> 00:25:00.277 on Runway 298 00:25:11.980 --> 00:25:16.740 you can upload a video here 299 00:25:16.740 --> 00:25:25.142 For example if I put the video I prepared 300 00:25:25.558 --> 00:25:27.656 Let's put Demo Assets here 301 00:25:33.448 --> 00:25:36.251 This is called the original input 302 00:25:36.321 --> 00:25:38.218 With the original input 303 00:25:40.020 --> 00:25:43.674 you can adjust the settings 304 00:25:48.031 --> 00:25:49.021 like style 305 00:25:54.160 --> 00:25:57.479 weight, and seed 306 00:25:57.479 --> 00:26:02.050 After adjusting them all and generate here 307 00:26:02.050 --> 00:26:09.000 you can get results according to these settings 308 00:26:09.761 --> 00:26:14.670 And these another very famous AI tool in Image to Video 309 00:26:14.670 --> 00:26:17.406 It is Stable Diffusion 310 00:26:17.911 --> 00:26:20.576 I think you would have heard of it a lot 311 00:26:24.120 --> 00:26:27.880 In Instagram or Youtube 312 00:26:27.880 --> 00:26:35.172 they are uploading a lot of content 313 00:26:36.370 --> 00:26:39.815 This one as well, 314 00:26:43.181 --> 00:26:45.352 if you put an image here 315 00:26:49.381 --> 00:26:50.991 if generates a video 316 00:26:51.110 --> 00:26:53.246 So in Advanced Options 317 00:26:57.949 --> 00:26:59.864 if you just do Generate 318 00:27:00.250 --> 00:27:02.920 But for example when you see 319 00:27:02.920 --> 00:27:08.188 when generating, Stable Diffusion might 320 00:27:09.050 --> 00:27:11.479 take a very long ime 321 00:27:11.479 --> 00:27:22.281 Because this one, it doesn't use my computer's GPU 322 00:27:23.290 --> 00:27:26.456 and uses Stable Diffusion's GPU instead 323 00:27:27.090 --> 00:27:32.003 so if there are a lot of works loaded there 324 00:27:32.161 --> 00:27:37.227 mine might be pushed back and take a long time to generate 325 00:27:37.761 --> 00:27:40.079 but the generation itself is very fast 326 00:27:40.079 --> 00:27:45.862 That is because it is using their computer 327 00:27:46.120 --> 00:27:49.264 If my computer was a very good computer 328 00:27:51.660 --> 00:27:52.821 it will be generated very quickly 329 00:27:52.821 --> 00:27:57.082 But if my computer has a 330 00:27:57.309 --> 00:28:00.557 very poor graphic card 331 00:28:01.250 --> 00:28:03.719 then this Stable Diffusion might be very helpful 332 00:28:03.719 --> 00:28:05.719 because it uses another 333 00:28:05.719 --> 00:28:09.353 very good computer's network 334 00:28:09.640 --> 00:28:16.221 So it does have the disadvantage of taking a long time due to the queue 335 00:28:16.221 --> 00:28:19.187 For example, if you take a look at the examples here 336 00:28:23.860 --> 00:28:27.515 like this, animation of eyes blinking 337 00:28:29.010 --> 00:28:29.875 Let's see other ones as well 338 00:28:31.281 --> 00:28:33.079 The left is the picture 339 00:28:33.079 --> 00:28:37.603 and the right is the generated video 340 00:28:38.811 --> 00:28:41.565 So in the back, the fire 341 00:28:42.070 --> 00:28:51.858 or with the panning the girl in the front is also moving a little 342 00:28:59.600 --> 00:29:10.970 It automatically adjusts the changed looks due to the 343 00:29:10.970 --> 00:29:13.554 change in angle of the face 344 00:29:13.930 --> 00:29:17.743 so there are some awkward parts when the girl turns her face 345 00:29:19.671 --> 00:29:23.914 but the results by now 346 00:29:25.300 --> 00:29:29.188 don't seem to be too bad 347 00:29:30.337 --> 00:29:32.662 So some people say 348 00:29:32.741 --> 00:29:37.427 'these videos made with AI now 349 00:29:38.239 --> 00:29:41.300 are equivalent to the black and white movies 350 00:29:41.300 --> 00:29:45.537 before color movies' 351 00:29:46.151 --> 00:29:52.608 I believe this was said in the video 352 00:29:52.608 --> 00:29:55.529 making Sora 353 00:29:56.301 --> 00:29:58.319 Considering that speed 354 00:29:58.319 --> 00:30:00.968 Considering how we came from black and white movies to VFX movies 355 00:30:01.720 --> 00:30:05.265 like Avatar and such 356 00:30:05.631 --> 00:30:11.669 movies made with AI will be possible very soon 357 00:30:11.669 --> 00:30:13.140 Quicker than our expectations 358 00:30:15.081 --> 00:30:17.696 It is thought to be possible 359 00:30:18.320 --> 00:30:20.272 And then Dall-E 360 00:30:20.401 --> 00:30:22.640 As you all know 361 00:30:22.640 --> 00:30:27.689 this is a image generation tool from OpenAI 362 00:30:28.630 --> 00:30:31.863 But if you go to Dall-E 363 00:30:32.140 --> 00:30:37.064 it is out of service and it's waiting for the next version 364 00:30:37.520 --> 00:30:42.363 I assume that that will be Sora 365 00:30:43.610 --> 00:30:47.347 But it is currently out of service 366 00:30:49.040 --> 00:30:54.800 So Runway, Pika, and Stable Diffusion 367 00:30:54.800 --> 00:30:57.962 did write the prompts 368 00:30:58.170 --> 00:31:04.881 but still generated videos 369 00:31:05.020 --> 00:31:07.985 based on the picture, drawing, or video 370 00:31:08.401 --> 00:31:10.325 This Dall-E 371 00:31:11.791 --> 00:31:17.694 generates video only with prompts 372 00:31:18.140 --> 00:31:21.739 or generates images 373 00:31:22.878 --> 00:31:27.439 Stable Diffusion also has that function 374 00:31:27.439 --> 00:31:29.478 If you visit the website 375 00:31:31.201 --> 00:31:35.292 you can choose from many options 376 00:31:35.569 --> 00:31:37.628 and generate one 377 00:31:39.400 --> 00:31:42.319 Usually Stable Diffusion 378 00:31:42.319 --> 00:31:44.955 or Midjourney 379 00:31:45.470 --> 00:31:48.426 use Discord a lot 380 00:31:49.594 --> 00:31:53.108 If you join Discord 381 00:31:53.850 --> 00:31:57.023 and make your AI window 382 00:31:58.281 --> 00:32:06.520 put a slash and click 'imagine' 383 00:32:06.520 --> 00:32:09.585 then write a prompt 384 00:32:10.050 --> 00:32:15.216 Then this 'imagine' means using the Midjourney bot 385 00:32:16.058 --> 00:32:21.844 so writing 'imagine' and then the prompt 386 00:32:23.161 --> 00:32:26.230 this gets automatically connected to Midjourney 387 00:32:27.396 --> 00:32:29.671 to create an image in Discord 388 00:32:29.671 --> 00:32:30.579 For example 389 00:32:32.520 --> 00:32:33.555 For example 390 00:32:36.090 --> 00:32:41.430 chest shot of asian 391 00:32:55.073 --> 00:33:00.630 chest shot of asian guy in seoul city 392 00:33:00.630 --> 00:33:03.982 For example if I put the prompt like this 393 00:33:06.239 --> 00:33:09.472 it will automatically connect with Midjourney 394 00:33:10.610 --> 00:33:11.878 and generate one 395 00:33:14.511 --> 00:33:20.932 Of course there are many AI tools made in Korea 396 00:33:22.120 --> 00:33:25.168 but for now, tools from the US 397 00:33:25.970 --> 00:33:29.974 have much better performance that the ones from Korea 398 00:33:31.331 --> 00:33:33.841 and since they are from the US 399 00:33:34.831 --> 00:33:37.081 the prompt also must be written in English 400 00:33:37.081 --> 00:33:40.430 to get a good result 401 00:33:41.301 --> 00:33:44.079 So if you write it in Korean 402 00:33:44.079 --> 00:33:46.161 the result might 403 00:33:47.270 --> 00:33:52.285 not be as good as you expected 404 00:33:52.740 --> 00:33:56.810 And also how you write the prompt 405 00:33:56.810 --> 00:33:58.837 really changes the result 406 00:33:59.461 --> 00:34:01.800 So constantly checking while writing 407 00:34:01.800 --> 00:34:04.839 and checking what the write prompt for the tool is 408 00:34:04.839 --> 00:34:07.436 and turning that into a list 409 00:34:07.941 --> 00:34:11.926 so that you can use that when 410 00:34:12.887 --> 00:34:15.119 making other shots 411 00:34:15.119 --> 00:34:16.995 I think that should be how it should be practiced 412 00:34:17.431 --> 00:34:20.359 So normally when using Midjourney 413 00:34:20.359 --> 00:34:24.429 It generates four images 414 00:34:25.834 --> 00:34:26.919 They all have different vibes 415 00:34:26.919 --> 00:34:30.109 So you can choose one among these 416 00:34:31.030 --> 00:34:36.204 and take it again to Runway or Stable Diffusion 417 00:34:36.600 --> 00:34:38.243 to create another shot 418 00:34:38.441 --> 00:34:40.577 and then edit them to make a video 419 00:34:41.191 --> 00:34:44.000 These processes are needed 420 00:34:44.941 --> 00:34:51.501 Up until now we learned about image generation methods 421 00:34:52.679 --> 00:34:58.595 centered around production in animation production 422 00:34:58.991 --> 00:35:03.959 apart from these, as you can see in our materials 423 00:35:03.959 --> 00:35:06.139 there are a lot of tools 424 00:35:07.851 --> 00:35:12.441 and these will be the tools 425 00:35:12.441 --> 00:35:16.593 used in the production when making animations 426 00:35:18.870 --> 00:35:24.726 Then using these, when you produce a video 427 00:35:25.251 --> 00:35:28.321 you'll be able to use many of them 428 00:35:28.321 --> 00:35:31.860 as you can see in the examples 429 00:35:32.994 --> 00:35:35.794 Prompt to Video Runway Gen 2 Pika Labs Image to Video Runway Gen 2 Pika Labs Stable Video Diffusion 430 00:35:35.794 --> 00:35:38.580 Video to Stylized Video Runway Gen 1 431 00:35:38.580 --> 00:35:42.030 Text to Image Tools Stable Diffusion Dall-E Friefly Midjourney Voice Over / Dubbing Elevenlabs Lovo 432 00:35:42.030 --> 00:35:45.489 Ideas and Scriptwriting Bard ChatGPT Translation Heygen Elevenlabs 433 00:35:45.726 --> 00:35:50.360 To used more diversely 434 00:35:50.360 --> 00:35:53.604 would be a bit difficult for now 435 00:35:54.921 --> 00:35:58.933 But still, I hope you can try producing 436 00:35:59.181 --> 00:36:02.120 using the tools I introduced 437 00:36:02.120 --> 00:36:02.790 Thank you 438 00:36:05.271 --> 00:36:12.558 Animation storytelling using AI Characteristics of AI storytelling New stories Creative stories Diverse stories 439 00:36:12.558 --> 00:36:16.308 2. AI tools that can be used in animation production Characteristics of AI tools It is recommended to try free trials and then subscribing to appropriate tools on websites since purchasing or downloading is not available 440 00:36:16.308 --> 00:36:17.039 Prompt must be written in English to get good results Types of AI tools Prompt to Video: Runway Gen 2, Pika Labs 441 00:36:17.039 --> 00:36:17.811 Video to Stylized Video: Runway Gen 1 Image to Video:Runway Gen 2, Pika Labs, Stable Video Diffusion 442 00:36:17.811 --> 00:36:18.663 Text to Image Tools: Stable Diffusion, Dall-E, Friefly, Midjourney 443 00:36:18.663 --> 00:36:19.301 Ideas and Scriptwriting: Bard, ChatGPT Voice Over / Dubbing : Elevenlabs, Lovo 444 00:36:19.301 --> 00:36:20.043 Translation: Heygen, Elevenlabs