WEBVTT NOTE This file was generated by switchtitle 1 00:00:13.500 --> 00:00:15.625 This is Self-Directed Research. Our 2 00:00:15.625 --> 00:00:17.125 hosts, James and Amos, 3 00:00:17.125 --> 00:00:18.041 entertain each other with 4 00:00:18.041 --> 00:00:20.000 weekly hyper-focused technical deep 5 00:00:20.000 --> 00:00:21.375 dives. Amos continues 6 00:00:21.375 --> 00:00:22.708 on the arc this week with 7 00:00:22.791 --> 00:00:25.666 "How Far Can You Take OpenFX?" Make sure 8 00:00:25.666 --> 00:00:26.500 to like, follow, and 9 00:00:26.500 --> 00:00:27.916 subscribe wherever you find us, 10 00:00:27.916 --> 00:00:30.708 and visit sdr-podcast.com/episodes 11 00:00:30.708 --> 00:00:31.083 for all the 12 00:00:31.083 --> 00:00:32.750 presentations, videos, show notes, 13 00:00:32.750 --> 00:00:35.041 and transcripts. Also, mark your 14 00:00:35.041 --> 00:00:36.333 calendars! Self-Directed 15 00:00:36.333 --> 00:00:37.583 Research will be recording live 16 00:00:37.750 --> 00:00:40.833 at EuroRust in Paris and online October 17 00:00:40.833 --> 00:00:42.208 9th and 10th, 2025. 18 00:00:42.750 --> 00:00:44.333 Don't miss it! And as usual, 19 00:00:44.791 --> 00:00:46.375 stay tuned to the end to hear more about 20 00:00:46.375 --> 00:00:48.416 Depot, the sponsor of this episode. 21 00:00:54.458 --> 00:00:56.458 So I've been up all night. 22 00:00:56.458 --> 00:00:57.208 No, that's not true. 23 00:00:57.583 --> 00:00:58.791 Trying to figure out what 24 00:00:58.791 --> 00:01:00.500 to bring to show the class. 25 00:01:01.166 --> 00:01:03.166 And I thought it might be fun if we did 26 00:01:03.166 --> 00:01:05.875 the freaking news, but I didn't. 27 00:01:06.083 --> 00:01:06.416 I don't know. 28 00:01:06.583 --> 00:01:07.500 It kept me up. 29 00:01:07.750 --> 00:01:10.833 So I opened my iPhone Notes app just to 30 00:01:10.833 --> 00:01:12.583 list all the things I could talk about. 31 00:01:12.958 --> 00:01:15.708 And I decided to use the research that I 32 00:01:15.708 --> 00:01:18.958 did yesterday to kind of close the arc of 33 00:01:18.958 --> 00:01:21.625 me going deeper and deeper into rabbit 34 00:01:21.625 --> 00:01:25.916 holes when it comes to making videos the 35 00:01:25.916 --> 00:01:27.125 silliest possible way. 36 00:01:27.458 --> 00:01:27.958 James, are you excited? 37 00:01:28.208 --> 00:01:28.666 I am. 38 00:01:29.083 --> 00:01:30.833 I'm glad I was using Bluetooth headphones 39 00:01:30.833 --> 00:01:32.333 because I did wander away to get my 40 00:01:32.333 --> 00:01:33.625 fidget toy, but I did 41 00:01:33.625 --> 00:01:34.500 hear the whole intro. 42 00:01:34.750 --> 00:01:35.833 Good. Good, good, good. 43 00:01:36.208 --> 00:01:37.666 So today's episode is called 44 00:01:37.666 --> 00:01:39.791 How Far Can You Take OpenFX? 45 00:01:40.250 --> 00:01:42.416 Subtitle, can you embed a full motion 46 00:01:42.416 --> 00:01:43.500 design suite with it? 47 00:01:43.500 --> 00:01:45.458 But I would like to manage everyone's 48 00:01:45.458 --> 00:01:47.291 expectations, including my own. 49 00:01:47.541 --> 00:01:51.000 And it's that I only put like 50 00:01:51.000 --> 00:01:52.916 one day of research into this. 51 00:01:53.125 --> 00:01:54.375 For folks who haven't listened to your 52 00:01:54.375 --> 00:01:57.166 last episode where we were introduced to 53 00:01:57.166 --> 00:01:59.375 what OpenFX, do you want to remind the 54 00:01:59.375 --> 00:02:01.708 audience what OpenFX is and where you 55 00:02:01.708 --> 00:02:03.458 start this rabbit hole dive in? 56 00:02:03.708 --> 00:02:05.250 James, ye of little faith. 57 00:02:05.666 --> 00:02:06.916 Well, I've got to ask for the audience 58 00:02:06.916 --> 00:02:09.458 Previously, on Self-Directed Research. 59 00:02:09.958 --> 00:02:11.125 Oh, love it. 60 00:02:11.333 --> 00:02:11.708 Love it. 61 00:02:11.958 --> 00:02:13.375 So first of all, as always, you can get 62 00:02:13.375 --> 00:02:16.250 the slides on SDR dash podcast dot com 63 00:02:16.250 --> 00:02:19.375 slash episodes or by watching the episode 64 00:02:19.375 --> 00:02:20.666 on YouTube or Spotify. 65 00:02:21.166 --> 00:02:21.916 We have video on Spotify. 66 00:02:23.416 --> 00:02:25.750 And for those of you who don't know, I 67 00:02:25.750 --> 00:02:26.791 make YouTube videos. 68 00:02:27.375 --> 00:02:29.291 And on the slide, we can currently see 69 00:02:29.291 --> 00:02:30.250 what my videos look like. 70 00:02:30.250 --> 00:02:31.750 It's mostly a talking 71 00:02:31.750 --> 00:02:34.083 head, mine, in front of code. 72 00:02:34.291 --> 00:02:35.250 Sometimes there's a green screen 73 00:02:35.250 --> 00:02:36.125 involved, but we don't 74 00:02:36.125 --> 00:02:37.083 need to get into that. 75 00:02:37.083 --> 00:02:38.375 The important part is how do we render 76 00:02:38.375 --> 00:02:40.791 that code and how do we like animated and 77 00:02:40.791 --> 00:02:42.333 highlight parts of it and whatnot? 78 00:02:42.375 --> 00:02:44.458 And I've gone through several different 79 00:02:44.458 --> 00:02:45.958 techniques to do that. 80 00:02:46.208 --> 00:02:48.625 It all started with screenshots because 81 00:02:48.625 --> 00:02:49.458 the code blocks are 82 00:02:49.458 --> 00:02:50.208 already on my website. 83 00:02:50.500 --> 00:02:50.708 Right. 84 00:02:50.875 --> 00:02:53.083 I my website is also all custom. 85 00:02:53.250 --> 00:02:54.708 It does syntax highlighting. 86 00:02:55.166 --> 00:02:57.291 So they're already rendered exactly the 87 00:02:57.291 --> 00:02:58.250 way I would like to 88 00:02:58.250 --> 00:02:59.541 present them on my website. 89 00:02:59.791 --> 00:03:01.166 So ideally, I just take that 90 00:03:01.166 --> 00:03:02.291 and put that in the videos. 91 00:03:02.583 --> 00:03:04.958 And that's what I used to do is just use 92 00:03:04.958 --> 00:03:06.875 the Mac OS screenshot tool and do a 93 00:03:06.875 --> 00:03:07.875 rectangle screenshot 94 00:03:07.875 --> 00:03:09.208 and save that somewhere. 95 00:03:09.416 --> 00:03:12.000 Rename that to something meaningful. 96 00:03:12.541 --> 00:03:14.000 Make sure I don't lose it because that's 97 00:03:14.000 --> 00:03:14.958 the thing with 98 00:03:14.958 --> 00:03:15.958 DaVinci Resolve projects 99 00:03:15.958 --> 00:03:18.000 The cloud is a relative concept. 100 00:03:18.500 --> 00:03:21.041 And yeah, it just took a lot of time and 101 00:03:21.041 --> 00:03:23.208 it was hard to position them properly. 102 00:03:23.208 --> 00:03:25.458 If you go from one slide to the next, if 103 00:03:25.458 --> 00:03:27.541 you if you did the rectangle screenshot 104 00:03:27.541 --> 00:03:28.958 slightly different from 105 00:03:28.958 --> 00:03:30.791 the one before, it's funky. 106 00:03:31.041 --> 00:03:32.333 You can definitely see it in the 107 00:03:32.333 --> 00:03:33.666 transition from one to the other. 108 00:03:34.208 --> 00:03:36.125 The controls in the venture resolve to 109 00:03:36.125 --> 00:03:37.666 set the position and size 110 00:03:37.666 --> 00:03:39.916 of the text are very weird. 111 00:03:40.166 --> 00:03:40.708 I don't even know what 112 00:03:40.708 --> 00:03:42.125 they're based on, honestly. 113 00:03:42.791 --> 00:03:44.458 And, you know, it was not good. 114 00:03:44.458 --> 00:03:46.416 So I started looking for solutions on how 115 00:03:46.416 --> 00:03:47.791 to make this easier because I usually 116 00:03:47.791 --> 00:03:50.916 write an entire article first and then 117 00:03:50.916 --> 00:03:51.833 make a video out of it. 118 00:03:52.875 --> 00:03:54.000 So I have everything. 119 00:03:54.375 --> 00:03:57.541 So I can just open the article in a 120 00:03:57.541 --> 00:03:58.541 browser like Safari. 121 00:03:58.791 --> 00:04:00.375 And I noticed that in Safari in the dev 122 00:04:00.375 --> 00:04:01.666 tools, well, they call 123 00:04:01.666 --> 00:04:02.333 it the Web Inspector. 124 00:04:02.666 --> 00:04:04.458 But you can make transparent screenshots. 125 00:04:04.875 --> 00:04:07.083 And then I in the automating macOS 126 00:04:07.083 --> 00:04:09.708 episode of the season, we discovered that 127 00:04:09.708 --> 00:04:11.583 we can use Apple scripts to automate 128 00:04:11.583 --> 00:04:14.250 Safari, which means you can make those 129 00:04:14.250 --> 00:04:16.208 transparent screenshots automatically. 130 00:04:16.875 --> 00:04:18.666 And for one of the videos I had, like, I 131 00:04:18.666 --> 00:04:20.541 don't know, like 200 of them, blocks of 132 00:04:20.541 --> 00:04:22.625 code and diagrams and whatnot. 133 00:04:22.875 --> 00:04:24.000 So, you know, it gets 134 00:04:24.000 --> 00:04:25.291 very long to do it by hand. 135 00:04:25.750 --> 00:04:26.416 Which one was that? 136 00:04:26.750 --> 00:04:27.250 I forget. 137 00:04:27.833 --> 00:04:28.958 Honestly, I have. 138 00:04:29.208 --> 00:04:30.958 OK, I'm lost in my pipeline. 139 00:04:31.166 --> 00:04:32.375 To be honest, you probably could have 140 00:04:32.375 --> 00:04:33.208 said anything and it 141 00:04:33.208 --> 00:04:33.958 wouldn't surprise me. 142 00:04:34.458 --> 00:04:37.750 Yeah, I started making shorter pieces 143 00:04:37.750 --> 00:04:40.833 because it's just not nice to have a 50 144 00:04:40.833 --> 00:04:42.458 minute video to edit, I think 145 00:04:42.458 --> 00:04:45.166 some people here can relate. 146 00:04:45.166 --> 00:04:46.333 Now you're considering the 147 00:04:46.333 --> 00:04:48.083 niceness of your editors time. 148 00:04:48.083 --> 00:04:50.333 And so that just being your time, you go, 149 00:04:50.333 --> 00:04:52.333 well, I would I wouldn't inflict that on 150 00:04:52.333 --> 00:04:53.250 anyone, let alone 151 00:04:53.250 --> 00:04:54.375 someone who works with me. 152 00:04:54.875 --> 00:04:55.041 Yeah. 153 00:04:55.750 --> 00:04:57.250 I appreciate it. 154 00:04:57.583 --> 00:04:58.375 The past I don't know how 155 00:04:58.375 --> 00:04:59.708 many episodes have been like, 156 00:05:00.708 --> 00:05:01.833 said Amanda, who's been 157 00:05:01.833 --> 00:05:03.291 editing 50 minute episodes. 158 00:05:03.708 --> 00:05:04.416 Don't worry, Amanda, 159 00:05:04.541 --> 00:05:05.208 this one will be a short. 160 00:05:05.208 --> 00:05:05.541 Yeah, exactly. 161 00:05:05.708 --> 00:05:07.583 Only 44 slides or what is it? 162 00:05:07.875 --> 00:05:08.083 Yeah. 163 00:05:08.125 --> 00:05:08.333 Yeah. 164 00:05:08.375 --> 00:05:08.541 Yeah. 165 00:05:08.791 --> 00:05:09.125 Nice. 166 00:05:09.541 --> 00:05:11.000 I've aligned incentives for my videos 167 00:05:11.000 --> 00:05:12.208 because I paid my editor 168 00:05:12.208 --> 00:05:14.125 by minute of edited video. 169 00:05:14.458 --> 00:05:14.666 Nice. 170 00:05:15.125 --> 00:05:15.875 So I have an 171 00:05:15.875 --> 00:05:17.041 incentive to make them short. 172 00:05:17.333 --> 00:05:18.583 And then we discussed the rate and 173 00:05:18.583 --> 00:05:20.208 evolve and evolve it over time. 174 00:05:20.541 --> 00:05:22.000 But so that's what we 175 00:05:22.000 --> 00:05:23.333 discussed in automating Mac OS. 176 00:05:23.583 --> 00:05:24.333 You just use Apple 177 00:05:24.333 --> 00:05:25.750 script, which was very fun. 178 00:05:25.750 --> 00:05:26.875 There's on the slide. 179 00:05:26.875 --> 00:05:27.958 If you're watching the 180 00:05:27.958 --> 00:05:29.208 slide, you can see that there is 181 00:05:30.208 --> 00:05:34.458 not a mono space font, but a serif font. 182 00:05:34.875 --> 00:05:35.791 Actually, is it serif? 183 00:05:36.041 --> 00:05:36.833 I don't see serif. 184 00:05:36.833 --> 00:05:37.291 Is it sans? 185 00:05:37.875 --> 00:05:38.083 Whatever. 186 00:05:38.166 --> 00:05:39.875 It's that it's proportional for sure, 187 00:05:40.166 --> 00:05:41.541 which is the opposite of monospace. 188 00:05:41.541 --> 00:05:42.208 It's sans serif. 189 00:05:42.541 --> 00:05:44.416 There are no serifs on this font. 190 00:05:44.416 --> 00:05:46.250 I thought maybe the the lack of sleep 191 00:05:46.250 --> 00:05:47.750 made me the sans disappear. 192 00:05:48.125 --> 00:05:49.083 The serifs disappear. 193 00:05:50.250 --> 00:05:52.125 And another as the episode named 194 00:05:52.125 --> 00:05:53.291 Blackmagic Fusion is weird. 195 00:05:53.666 --> 00:05:57.250 I discussed my new and current setup. 196 00:05:57.333 --> 00:05:59.083 I've already reached a video with it. 197 00:05:59.083 --> 00:06:00.791 I forget which one, which involved 198 00:06:00.791 --> 00:06:05.416 reversing the Lua based formats that 199 00:06:05.416 --> 00:06:08.250 Blackmagic Fusion uses for node graphs 200 00:06:08.250 --> 00:06:10.458 and just kind of generating what it 201 00:06:10.458 --> 00:06:12.916 expects and generating a Text+ node 202 00:06:12.916 --> 00:06:15.791 with the character level styling modifier 203 00:06:15.791 --> 00:06:19.666 and doing syntax highlighting that way. 204 00:06:19.666 --> 00:06:21.708 And I even vibe coded a 205 00:06:21.708 --> 00:06:23.041 little preview on my website. 206 00:06:23.083 --> 00:06:24.916 So you press a button and then it opens 207 00:06:24.916 --> 00:06:27.708 this editor that uses "contenteditable" 208 00:06:28.125 --> 00:06:30.333 So you can remove the parts that you're 209 00:06:30.333 --> 00:06:31.458 we're not talking about in the video 210 00:06:31.458 --> 00:06:33.500 right now, you can like select a bit of 211 00:06:33.500 --> 00:06:34.666 text and press space. 212 00:06:34.666 --> 00:06:35.791 And it actually highlight 213 00:06:35.791 --> 00:06:37.541 this and dims everything else. 214 00:06:37.791 --> 00:06:39.166 It's really, really nice. 215 00:06:39.416 --> 00:06:42.083 It has a stand in for my face. 216 00:06:42.083 --> 00:06:44.291 I look super goofy again. 217 00:06:44.291 --> 00:06:45.375 Really, you should check the slides. 218 00:06:45.666 --> 00:06:46.333 That's half the fun. 219 00:06:47.750 --> 00:06:49.000 This is disappointing because this is one 220 00:06:49.000 --> 00:06:50.791 of the few website plugins that only 221 00:06:50.791 --> 00:06:51.791 works locally and you 222 00:06:51.791 --> 00:06:52.708 can't trigger it remotely. 223 00:06:52.833 --> 00:06:54.708 Because otherwise I'd love to like edit 224 00:06:54.708 --> 00:06:57.750 your page and then slap your face in on 225 00:06:57.750 --> 00:07:00.250 examples and be like you are making your 226 00:07:00.250 --> 00:07:02.291 own meme templates except for it's. 227 00:07:02.291 --> 00:07:04.916 I need a new patrons here for like like 228 00:07:04.916 --> 00:07:05.916 green screen pictures of me. 229 00:07:06.208 --> 00:07:06.500 Yeah, sure. 230 00:07:07.750 --> 00:07:09.208 Yeah, exactly. 231 00:07:09.208 --> 00:07:09.916 For evil people. 232 00:07:10.166 --> 00:07:12.583 I just saw yesterday on Reddit a post 233 00:07:12.583 --> 00:07:14.583 where some burger place lets you pay. 234 00:07:14.916 --> 00:07:16.166 The Veruca Salt special. 235 00:07:16.458 --> 00:07:18.500 You can pay a thousand dollars to 236 00:07:18.500 --> 00:07:20.125 get in front of the line and just. 237 00:07:20.625 --> 00:07:21.583 I want it now, Daddy. 238 00:07:21.875 --> 00:07:23.000 I think that's what it said. 239 00:07:23.500 --> 00:07:23.708 Exactly. 240 00:07:23.875 --> 00:07:24.041 Yeah. 241 00:07:24.333 --> 00:07:25.625 And I think that's something people 242 00:07:25.625 --> 00:07:26.625 should consider, like 243 00:07:26.625 --> 00:07:27.875 making tiers that are higher. 244 00:07:28.500 --> 00:07:30.083 So yeah, that reminds me of also someone 245 00:07:30.083 --> 00:07:31.916 said like on Twitter a long time ago, you 246 00:07:31.916 --> 00:07:33.125 know, when Twitter was still Twitter, 247 00:07:33.125 --> 00:07:34.958 they said like if someone were to pay me 248 00:07:34.958 --> 00:07:36.875 like 5000 a month, I would just stop 249 00:07:36.875 --> 00:07:37.833 posting indefinitely. 250 00:07:38.458 --> 00:07:39.250 And that's true. 251 00:07:41.000 --> 00:07:41.416 I do. 252 00:07:41.416 --> 00:07:42.958 Who was I'm going to have to go back and 253 00:07:42.958 --> 00:07:44.541 find that because I definitely remember 254 00:07:44.541 --> 00:07:46.416 that of the like it's someone we know. 255 00:07:46.625 --> 00:07:47.750 I don't I forget whom. 256 00:07:47.875 --> 00:07:49.958 Yeah, it was just I'll shut up, you know, 257 00:07:49.958 --> 00:07:51.375 for that check keeps clear. 258 00:07:51.375 --> 00:07:52.916 And I exactly I'll just shut the hell up. 259 00:07:53.166 --> 00:07:56.541 And it worked out beautifully on my blog. 260 00:07:56.541 --> 00:07:57.458 I have my articles that 261 00:07:57.458 --> 00:07:58.583 are exclusive for six months. 262 00:07:58.833 --> 00:08:00.333 Someone was like, hey, it's bad that 263 00:08:00.333 --> 00:08:01.333 exclusive for six months. 264 00:08:01.333 --> 00:08:02.666 And I was like, OK, you want to you want 265 00:08:02.666 --> 00:08:03.458 to pay to change that? 266 00:08:03.458 --> 00:08:03.958 I was like, OK, 267 00:08:05.541 --> 00:08:07.416 that was Depot, the sponsor of this episode. 268 00:08:07.875 --> 00:08:09.583 So a big thanks to them. 269 00:08:09.583 --> 00:08:10.458 Yeah, I was going to say it was our 270 00:08:10.458 --> 00:08:12.583 lovely sponsor, Depot, who is I know, who 271 00:08:12.583 --> 00:08:13.500 is sponsored this entire 272 00:08:13.500 --> 00:08:15.375 season and that article. 273 00:08:15.750 --> 00:08:17.000 Very generous and very cool platform. 274 00:08:17.000 --> 00:08:17.375 Check them out. 275 00:08:17.666 --> 00:08:17.791 Yeah. 276 00:08:17.791 --> 00:08:18.666 Listen to the end of the episode. 277 00:08:19.041 --> 00:08:19.250 Yeah. 278 00:08:19.458 --> 00:08:21.375 Back to my short presentation in the 279 00:08:21.375 --> 00:08:22.833 Black Magic Fusion is Weird episode. 280 00:08:23.291 --> 00:08:24.791 It's kind of a misnomer because it has 281 00:08:24.791 --> 00:08:26.666 this whole nice solution of generating. 282 00:08:26.666 --> 00:08:28.375 And I also talked about the other ways in 283 00:08:28.375 --> 00:08:30.333 which fusion can be extended, including 284 00:08:30.333 --> 00:08:32.416 fuses, which we learned come from the 285 00:08:32.416 --> 00:08:33.666 same word as fusion. 286 00:08:34.041 --> 00:08:35.458 We weren't sure during the episode, but 287 00:08:35.458 --> 00:08:37.541 in the in the show notes that Amanda put 288 00:08:37.541 --> 00:08:40.333 together, I saw that the etymology says 289 00:08:40.333 --> 00:08:42.083 that they are from the same word. 290 00:08:42.083 --> 00:08:42.541 Were they consistent? 291 00:08:42.833 --> 00:08:43.708 I saw one was like a 292 00:08:43.708 --> 00:08:45.916 bundle of twine, like "fuset" 293 00:08:45.958 --> 00:08:47.000 I think it was a French word. 294 00:08:47.250 --> 00:08:48.083 And then I don't remember 295 00:08:48.083 --> 00:08:50.083 where fusion came from, but 296 00:08:50.083 --> 00:08:53.291 "fusée" is a spaceship like a rocket ship. 297 00:08:53.291 --> 00:08:54.083 Oh, yeah. 298 00:08:54.083 --> 00:08:55.583 Maybe I'm probably pronouncing the French 299 00:08:55.583 --> 00:08:56.916 word very, very wrong. 300 00:08:57.541 --> 00:09:00.166 But yeah, my brain is too mushy to figure 301 00:09:00.166 --> 00:09:01.708 out which which French word you mean. 302 00:09:03.291 --> 00:09:04.916 We'll leave etymology for another day. 303 00:09:05.750 --> 00:09:05.958 Yeah. 304 00:09:06.083 --> 00:09:09.041 So fuses, which are a bits of Lua that you 305 00:09:09.041 --> 00:09:10.416 can hot reload and everything. 306 00:09:10.416 --> 00:09:12.291 It's a very nice development environment, 307 00:09:12.291 --> 00:09:13.625 which are fast enough 308 00:09:13.625 --> 00:09:15.375 because it's using a jit 309 00:09:15.416 --> 00:09:16.458 It's actually a LuaJIT 310 00:09:16.833 --> 00:09:18.375 Whenever people say Lua, they usually 311 00:09:18.375 --> 00:09:20.333 mean LuaJIT not always, but usually. 312 00:09:20.833 --> 00:09:22.166 And and OpenFX, I 313 00:09:22.166 --> 00:09:23.458 briefly touched on OpenFX. 314 00:09:23.458 --> 00:09:25.208 I was like, OK, it's a standard way to do 315 00:09:25.208 --> 00:09:28.208 video plug ins a little bit like AU 316 00:09:28.500 --> 00:09:30.833 and what was the one VST for audio? 317 00:09:31.333 --> 00:09:32.500 But this one is for video. 318 00:09:32.500 --> 00:09:34.916 And I said, you probably could bring in 319 00:09:34.916 --> 00:09:36.666 your own renderer and do 320 00:09:36.666 --> 00:09:37.958 exactly what you wanted. 321 00:09:38.250 --> 00:09:39.958 Yeah, one of the limitations we ran into 322 00:09:39.958 --> 00:09:41.958 by using back magic fusion's own text 323 00:09:41.958 --> 00:09:43.458 renderer is that it does not support 324 00:09:43.458 --> 00:09:44.708 color emoji, for example. 325 00:09:44.833 --> 00:09:47.041 And it does not support font fallbacks. 326 00:09:47.583 --> 00:09:49.333 So part of the difficulty for me in 327 00:09:49.333 --> 00:09:51.291 generating the graph is that if, for 328 00:09:51.291 --> 00:09:53.750 example, there is a nerd font symbol, it 329 00:09:53.750 --> 00:09:56.375 has to specify that this pan is using the 330 00:09:56.375 --> 00:09:57.625 nerd font symbol font. 331 00:09:57.916 --> 00:09:59.750 Everything else is in Berkeley Mono 332 00:09:59.750 --> 00:10:01.833 except for emojis, which are using Noto, 333 00:10:02.083 --> 00:10:03.416 Noto without color. 334 00:10:03.416 --> 00:10:04.666 monospace or the monochrome. 335 00:10:04.666 --> 00:10:05.708 Yeah, the monochrome. 336 00:10:05.708 --> 00:10:06.041 Yeah. 337 00:10:06.500 --> 00:10:07.875 And I went back and looked it up. 338 00:10:07.875 --> 00:10:08.875 And the word I thought 339 00:10:08.875 --> 00:10:10.416 was toast is actually tofu. 340 00:10:10.750 --> 00:10:12.583 And that's where the Noto Sans name 341 00:10:12.583 --> 00:10:14.583 comes from is "no tofu". 342 00:10:14.750 --> 00:10:16.125 Because it defines all of the 343 00:10:16.125 --> 00:10:17.666 symbols for all of the emojis. 344 00:10:17.958 --> 00:10:19.916 So you'll never get the tofu, which is 345 00:10:19.916 --> 00:10:21.333 that little rectangle that's either an 346 00:10:21.333 --> 00:10:23.541 exclamation point or has like the hex 347 00:10:23.541 --> 00:10:24.541 characters in there. 348 00:10:24.541 --> 00:10:26.166 I don't remember why it's called tofu, 349 00:10:26.375 --> 00:10:28.041 but that was the word I was trying to 350 00:10:28.041 --> 00:10:29.291 think of in that episode. 351 00:10:29.291 --> 00:10:31.250 The Unicode replacement character. 352 00:10:31.291 --> 00:10:31.583 Yeah. 353 00:10:32.000 --> 00:10:32.166 Yeah. 354 00:10:32.166 --> 00:10:33.416 Well, there's maybe two. 355 00:10:33.625 --> 00:10:34.333 There's one. 356 00:10:34.541 --> 00:10:35.000 Wow. 357 00:10:35.291 --> 00:10:35.916 No tofu. 358 00:10:35.916 --> 00:10:36.083 Yeah. 359 00:10:36.500 --> 00:10:38.208 But tofu was what I was thinking of when 360 00:10:38.208 --> 00:10:39.250 it shows you the like, 361 00:10:39.250 --> 00:10:40.375 I don't have the glyph. 362 00:10:40.583 --> 00:10:41.458 Here's some hex codes. 363 00:10:41.708 --> 00:10:42.458 Do you like that? 364 00:10:42.500 --> 00:10:44.291 I like learning about all this stuff, 365 00:10:44.666 --> 00:10:46.041 like going through editing the episode, 366 00:10:46.250 --> 00:10:47.708 collecting all the links and then being 367 00:10:47.708 --> 00:10:48.375 like, I don't know what 368 00:10:48.375 --> 00:10:49.416 you were talking about. 369 00:10:49.416 --> 00:10:50.791 And now next episode, 370 00:10:51.208 --> 00:10:53.875 learning about last week's episode. 371 00:10:54.125 --> 00:10:54.583 Very cool. 372 00:10:55.500 --> 00:10:57.291 I think you should feel free to cut in 373 00:10:57.291 --> 00:10:58.250 like add editors notes 374 00:10:58.250 --> 00:10:59.125 in the middle of episodes. 375 00:10:59.375 --> 00:10:59.958 Like I checked it. 376 00:11:00.208 --> 00:11:02.041 It's not true or something. 377 00:11:02.250 --> 00:11:03.125 They're still wrong. 378 00:11:03.458 --> 00:11:04.791 I was actually thinking about that. 379 00:11:04.791 --> 00:11:04.958 Yeah. 380 00:11:05.125 --> 00:11:06.416 But you know, who knows? 381 00:11:06.416 --> 00:11:07.250 Maybe in a future episode. 382 00:11:07.250 --> 00:11:07.583 We'll see. 383 00:11:07.958 --> 00:11:08.166 Yeah. 384 00:11:08.333 --> 00:11:10.500 So yeah, I mentioned OpenFX saying, yeah, 385 00:11:10.500 --> 00:11:11.791 you can you can make native plugins. 386 00:11:12.125 --> 00:11:13.833 You know, it's clearly overkill. 387 00:11:14.375 --> 00:11:16.291 Obviously, I'm not going to touch that. 388 00:11:16.583 --> 00:11:17.333 That would be crazy. 389 00:11:17.791 --> 00:11:18.958 So that's what I did next. 390 00:11:19.291 --> 00:11:19.708 Yeah, of course. 391 00:11:20.041 --> 00:11:20.583 Yeah, of course. 392 00:11:21.625 --> 00:11:23.375 Because because it's 393 00:11:23.375 --> 00:11:24.666 stuck in the back of my mind. 394 00:11:25.000 --> 00:11:26.041 There were some hints in there where 395 00:11:26.041 --> 00:11:27.500 you're like, oh, there's a crate. 396 00:11:27.750 --> 00:11:28.666 You know, I definitely 397 00:11:28.666 --> 00:11:30.250 exactly don't need that. 398 00:11:30.250 --> 00:11:30.916 I'll I'll definitely 399 00:11:30.916 --> 00:11:31.958 use the Lua for that. 400 00:11:31.958 --> 00:11:33.583 And it was one of those like, oh, no, 401 00:11:33.583 --> 00:11:34.958 it's this is Chekhov's 402 00:11:34.958 --> 00:11:36.333 gun in the last episode. 403 00:11:36.583 --> 00:11:37.500 Check Chekhov's crate. 404 00:11:37.500 --> 00:11:37.750 Yeah. 405 00:11:37.750 --> 00:11:39.625 That's what I was writing the slides. 406 00:11:39.625 --> 00:11:40.958 I was actually thinking of that line. 407 00:11:41.000 --> 00:11:42.250 Yeah, there was a crate for it. 408 00:11:42.250 --> 00:11:44.708 So how could I not look into it? 409 00:11:44.708 --> 00:11:45.250 Of course. 410 00:11:45.250 --> 00:11:47.083 I think the audience deserves that. 411 00:11:47.083 --> 00:11:48.750 I think that's what we're all here for. 412 00:11:48.958 --> 00:11:50.458 It's to watch me suffer. 413 00:11:50.458 --> 00:11:52.083 I was kind of immediately disappointed 414 00:11:52.083 --> 00:11:56.333 because the crates ofx-rs is written 415 00:11:56.333 --> 00:11:58.916 against nightly rust from six years ago. 416 00:11:59.125 --> 00:12:00.375 I didn't realize that I thought it was 417 00:12:00.375 --> 00:12:01.833 stable rust from six years ago, which 418 00:12:01.833 --> 00:12:02.666 would have been totally fine. 419 00:12:02.875 --> 00:12:04.791 I think maybe maybe a couple of fixes, 420 00:12:04.791 --> 00:12:08.000 but like unless something was wrong in 421 00:12:08.000 --> 00:12:10.041 terms of soundness, I 422 00:12:10.041 --> 00:12:11.041 think we would have been fine. 423 00:12:11.583 --> 00:12:13.000 But nightly from six years ago, I was 424 00:12:13.000 --> 00:12:16.041 using two features that just have changed 425 00:12:16.041 --> 00:12:17.625 or not there anymore, whatever. 426 00:12:18.041 --> 00:12:19.791 I got over 200 errors trying 427 00:12:19.791 --> 00:12:20.875 to compile the example code. 428 00:12:21.291 --> 00:12:23.041 So I thought that would be the end of it. 429 00:12:23.041 --> 00:12:24.208 That I told you it would be short. 430 00:12:24.500 --> 00:12:25.875 Yeah, I'm trying to like 2019 rust. 431 00:12:26.208 --> 00:12:27.666 It's not as far as it could be. 432 00:12:27.833 --> 00:12:29.125 We're like post 2018, 433 00:12:29.125 --> 00:12:30.583 which is a pretty big step. 434 00:12:30.583 --> 00:12:31.166 But yeah. 435 00:12:31.625 --> 00:12:32.041 Yeah. 436 00:12:32.041 --> 00:12:33.333 Six years of nightly is a stretch. 437 00:12:33.791 --> 00:12:34.083 Yeah. 438 00:12:34.291 --> 00:12:38.750 But then I found NTSC-RS, which is a port 439 00:12:38.750 --> 00:12:41.791 of NTSC-QT to rust 440 00:12:41.791 --> 00:12:43.833 and SIMD by Valadaptive. 441 00:12:44.208 --> 00:12:44.833 And among other 442 00:12:44.833 --> 00:12:47.458 things, it's 20 times faster. 443 00:12:47.666 --> 00:12:48.750 Yeah, it uses SIMD and 444 00:12:48.750 --> 00:12:49.708 concurrency and stuff. 445 00:12:49.708 --> 00:12:50.541 It's an analog. 446 00:12:50.875 --> 00:12:52.083 They have a very pretty website. 447 00:12:52.333 --> 00:12:52.750 It does. 448 00:12:53.125 --> 00:12:55.625 It's a VHS effect for your 449 00:12:55.625 --> 00:12:57.166 video filter, essentially. 450 00:12:57.416 --> 00:12:59.500 And it uses what does the website say? 451 00:13:00.083 --> 00:13:02.166 Uses algorithms that model how NTSC 452 00:13:02.166 --> 00:13:03.291 transmission and VHS 453 00:13:03.291 --> 00:13:05.333 encoding actually work. 454 00:13:05.333 --> 00:13:06.541 I imagine that is the next word. 455 00:13:06.791 --> 00:13:08.500 Ooh, someone was asking me about this on 456 00:13:08.500 --> 00:13:10.333 BlueSky because there's shaders that 457 00:13:10.333 --> 00:13:12.291 I've seen for terminals that give you 458 00:13:12.291 --> 00:13:14.250 that like CRT glow and stuff like that. 459 00:13:14.250 --> 00:13:15.875 But I'm not surprised that especially 460 00:13:15.875 --> 00:13:18.750 with all of the back rooms and like 461 00:13:18.750 --> 00:13:21.375 cassette futurism stuff that's going on, 462 00:13:21.375 --> 00:13:24.291 that there's not like super shaded CRTs 463 00:13:24.291 --> 00:13:27.041 and VHS things going on these days. 464 00:13:27.291 --> 00:13:27.500 True. 465 00:13:27.791 --> 00:13:28.541 True, true, true, true. 466 00:13:28.541 --> 00:13:31.625 That one is all CPU as far as I can tell, 467 00:13:31.625 --> 00:13:33.250 but it's I don't know. 468 00:13:33.250 --> 00:13:34.791 I think it's an old effect. 469 00:13:34.791 --> 00:13:36.083 Yeah, ported to rust. 470 00:13:36.500 --> 00:13:38.708 And the code place is very versatile 471 00:13:38.708 --> 00:13:39.916 because it has a standalone 472 00:13:39.916 --> 00:13:42.166 application with egui controls. 473 00:13:42.166 --> 00:13:43.416 I don't know if you recognize the 474 00:13:43.416 --> 00:13:45.583 egui-ness of the of the GUI. 475 00:13:45.583 --> 00:13:45.833 Definitely. 476 00:13:46.333 --> 00:13:47.208 Yeah, it could be. 477 00:13:47.458 --> 00:13:48.166 What's the other one? 478 00:13:48.166 --> 00:13:49.333 What's the C++ one? 479 00:13:49.583 --> 00:13:51.500 The one used in all the video games. 480 00:13:52.125 --> 00:13:52.500 Imgui. 481 00:13:52.791 --> 00:13:53.750 Oh, yeah, Dear Imgui. 482 00:13:53.958 --> 00:13:55.125 Yeah, Dear Dear Imgui. 483 00:13:55.125 --> 00:13:55.333 Exactly. 484 00:13:55.541 --> 00:13:57.125 I could be that, but I spend too much 485 00:13:57.125 --> 00:13:59.083 time with egui making demos. 486 00:13:59.375 --> 00:14:00.583 So it's available as a standalone. 487 00:14:01.000 --> 00:14:02.541 It's available as a CLI. 488 00:14:02.916 --> 00:14:04.125 It's available as an After 489 00:14:04.125 --> 00:14:05.208 Effects and Premiere plugin. 490 00:14:05.625 --> 00:14:07.750 And as an OpenFX plugin. 491 00:14:08.541 --> 00:14:09.583 And are we getting back into those 492 00:14:09.583 --> 00:14:10.875 plug-in world like we were talking about 493 00:14:10.875 --> 00:14:12.541 with the audio ones where it's like you 494 00:14:12.541 --> 00:14:13.958 have purchased a DLL. 495 00:14:14.375 --> 00:14:14.791 Congratulations. 496 00:14:15.250 --> 00:14:16.375 It runs in your environment. 497 00:14:16.625 --> 00:14:16.958 Kind of. 498 00:14:16.958 --> 00:14:17.958 Is it something much more 499 00:14:17.958 --> 00:14:19.625 specific to this platform? 500 00:14:20.291 --> 00:14:20.708 Kind of. 501 00:14:20.708 --> 00:14:22.291 Actually, so NTSC RS is open source. 502 00:14:22.708 --> 00:14:22.916 Okay. 503 00:14:23.083 --> 00:14:25.375 So actually they use cargo-xtask to 504 00:14:25.375 --> 00:14:27.875 build the plugin, but it's just so it can 505 00:14:27.875 --> 00:14:30.416 generate that little bundle, which looks 506 00:14:30.416 --> 00:14:31.750 like a Mac OS app bundle. 507 00:14:31.750 --> 00:14:33.458 It has a contents directory with an 508 00:14:33.458 --> 00:14:35.958 Info.plist in there, a MacOS directory 509 00:14:35.958 --> 00:14:38.500 with a .ofx, which is a shared library, 510 00:14:38.500 --> 00:14:39.333 resources directory 511 00:14:39.333 --> 00:14:41.583 with an icon of the plugin. 512 00:14:41.958 --> 00:14:44.125 And the info.plist is just XML. 513 00:14:44.458 --> 00:14:45.708 There's one of the plist formats. 514 00:14:45.708 --> 00:14:46.750 There's a binary one and 515 00:14:46.750 --> 00:14:48.041 there's a third one, I forget. 516 00:14:48.291 --> 00:14:49.250 Like text but not XML. 517 00:14:49.500 --> 00:14:50.083 I don't know James, 518 00:14:50.083 --> 00:14:50.958 does that ring a bell? 519 00:14:51.208 --> 00:14:51.791 Yeah, I gotcha. 520 00:14:51.791 --> 00:14:53.333 It's exactly the same kind of thing that 521 00:14:53.333 --> 00:14:54.958 I had to put together for the Postation 522 00:14:54.958 --> 00:14:56.833 bundle and for when you're 523 00:14:56.833 --> 00:14:57.958 shipping a binary on Mac OS. 524 00:14:58.166 --> 00:14:59.583 So it definitely rings true. 525 00:14:59.583 --> 00:15:00.958 Yeah, it looks a lot 526 00:15:00.958 --> 00:15:03.000 like it, but it's not .app. 527 00:15:03.083 --> 00:15:04.458 It's .ofx.bundle. 528 00:15:04.833 --> 00:15:05.666 It's weird. 529 00:15:06.375 --> 00:15:07.458 It's a folder that's 530 00:15:07.458 --> 00:15:08.791 pretending not to be a folder. 531 00:15:09.208 --> 00:15:11.041 And yeah, so the reason there's an 532 00:15:11.041 --> 00:15:13.458 xtask, which is kind of turning Cargo into 533 00:15:13.458 --> 00:15:15.208 make, you know, you can have different 534 00:15:15.208 --> 00:15:17.791 targets and it's "cargo xtask 535 00:15:17.791 --> 00:15:19.833 build-ofx-plugin" is because they want to 536 00:15:19.833 --> 00:15:20.708 make a universal build. 537 00:15:20.958 --> 00:15:22.125 I don't know if that's 538 00:15:22.125 --> 00:15:24.041 built into rustc yet. 539 00:15:24.416 --> 00:15:26.791 Nowadays, you make a binary for x86-64 540 00:15:26.791 --> 00:15:30.250 and for aarch64, but in the past you did it 541 00:15:30.250 --> 00:15:32.375 for like, well, Intel 32-bit and Intel 542 00:15:32.375 --> 00:15:34.208 64-bit, you put that in the same binary 543 00:15:34.375 --> 00:15:35.875 or a PowerPC plus until 544 00:15:35.875 --> 00:15:37.375 32-bit, even further back. 545 00:15:37.666 --> 00:15:39.958 So like any kind of transition that Apple 546 00:15:39.958 --> 00:15:42.083 has done, like Mac OS has done, you could 547 00:15:42.083 --> 00:15:44.750 have a transition period where you've 548 00:15:44.750 --> 00:15:46.791 shipped both binaries in the same thing. 549 00:15:46.791 --> 00:15:48.791 And it's still happening today because 550 00:15:48.791 --> 00:15:50.541 not everyone has Apple 551 00:15:50.541 --> 00:15:52.083 Silicon Mac, an ARM Mac. 552 00:15:52.333 --> 00:15:53.833 If you go on Visual Studio Code webpage, 553 00:15:54.041 --> 00:15:55.208 for example, there's a 554 00:15:55.208 --> 00:15:57.000 single download for Mac OS button. 555 00:15:57.000 --> 00:15:58.666 But if you download that and you use 556 00:15:58.666 --> 00:16:01.291 lipo-info, then you see that there's two 557 00:16:01.291 --> 00:16:03.958 architectures in there, x86-64 and arm64. 558 00:16:04.041 --> 00:16:05.333 I think even, I can't remember if it was 559 00:16:05.333 --> 00:16:08.208 Android or iOS, like for the iPhones and 560 00:16:08.208 --> 00:16:09.416 things like that, where they were doing 561 00:16:09.416 --> 00:16:11.000 that as well, because there's different 562 00:16:11.000 --> 00:16:13.250 generations of the processors that they 563 00:16:13.250 --> 00:16:15.500 use instead of doing the like, hey, we're 564 00:16:15.500 --> 00:16:18.125 compiling for the oldest compatible X86 565 00:16:18.125 --> 00:16:19.416 version and things like that. 566 00:16:19.416 --> 00:16:20.833 There's the option to do multiple 567 00:16:20.833 --> 00:16:22.375 compilations for different target 568 00:16:22.375 --> 00:16:24.500 platforms so that you can get either like 569 00:16:24.500 --> 00:16:27.791 the operating system platform level or 570 00:16:27.791 --> 00:16:30.291 like the CPU for Apple where there's much 571 00:16:30.291 --> 00:16:32.541 more control over like, hey, this works 572 00:16:32.541 --> 00:16:34.041 on these versions of phones and later, 573 00:16:34.041 --> 00:16:35.208 which means you have all of these 574 00:16:35.208 --> 00:16:37.833 processor extensions available for more 575 00:16:37.833 --> 00:16:39.000 performance and stuff like that. 576 00:16:39.000 --> 00:16:40.291 I wouldn't be surprised if both of them 577 00:16:40.291 --> 00:16:41.416 were doing that, but I guess it's just 578 00:16:41.416 --> 00:16:42.375 how you do the packaging 579 00:16:42.375 --> 00:16:43.958 and the app store handles it. 580 00:16:43.958 --> 00:16:46.125 It's possible, but given 581 00:16:46.125 --> 00:16:47.958 that these would be mobile apps. 582 00:16:48.583 --> 00:16:50.416 This is an aside, I will, yeah. 583 00:16:50.458 --> 00:16:53.791 I think the equation changes a little bit 584 00:16:53.791 --> 00:16:55.625 because who cares if VS Code is a few 585 00:16:55.625 --> 00:16:57.375 megabytes larger, but on 586 00:16:57.375 --> 00:16:59.125 the phone, I don't know. 587 00:16:59.125 --> 00:17:01.500 I know there's like SDK target levels for 588 00:17:01.500 --> 00:17:04.166 sure, and I know those are hard to handle 589 00:17:04.166 --> 00:17:05.958 properly, and those are also 590 00:17:05.958 --> 00:17:08.375 a thing on macOS, but yeah. 591 00:17:08.750 --> 00:17:09.958 But yeah, universal binaries. 592 00:17:10.291 --> 00:17:11.875 I remember them, like you said, from X86 593 00:17:11.875 --> 00:17:13.166 and ARM, but they 594 00:17:13.166 --> 00:17:14.333 continue to be relevant. 595 00:17:15.166 --> 00:17:18.083 If you want to save the bytes, you can go 596 00:17:18.083 --> 00:17:21.583 to that detailed download page and have 597 00:17:21.583 --> 00:17:23.958 like the Intel or Apple Silicon build 598 00:17:23.958 --> 00:17:26.791 separately and all the, oh yeah, for 599 00:17:26.791 --> 00:17:27.625 Linux, you get the 600 00:17:27.625 --> 00:17:30.458 ARM32 builds for like, what? 601 00:17:30.875 --> 00:17:32.458 What's ARM32 these days? 602 00:17:33.500 --> 00:17:34.458 Original Raspberry Pis. 603 00:17:35.083 --> 00:17:37.291 Raspberry Pi was 32-bit for a long while, 604 00:17:37.291 --> 00:17:39.041 and then it did a very weird thing where 605 00:17:39.041 --> 00:17:41.500 the kernel was 64-bit, but the user space 606 00:17:41.500 --> 00:17:44.041 was 32-bit, which was like a 607 00:17:44.041 --> 00:17:45.416 compat thing so that they didn't 608 00:17:45.416 --> 00:17:46.583 have to-- Because of RAM usage? 609 00:17:47.041 --> 00:17:49.166 Like 32-bit pointers or, okay. 610 00:17:49.666 --> 00:17:51.708 For compat, like if people had, yeah, so 611 00:17:51.708 --> 00:17:53.916 if people had compiled some binary 612 00:17:53.916 --> 00:17:57.000 application for 32-bit, you need a 32-bit 613 00:17:57.000 --> 00:17:58.458 user space to continue to run that, 614 00:17:58.458 --> 00:17:59.625 because otherwise you'd have to recompile 615 00:17:59.625 --> 00:18:02.416 to run on a 64-bit user space, but you 616 00:18:02.416 --> 00:18:03.666 might want your operating system to be 617 00:18:03.666 --> 00:18:07.083 able to handle more than the two, or be 618 00:18:07.083 --> 00:18:09.083 able to handle more RAM, or at least have 619 00:18:09.083 --> 00:18:10.750 extended RAM range, even if 620 00:18:10.750 --> 00:18:11.958 it's virtual memory range. 621 00:18:12.291 --> 00:18:13.125 So for a while they were doing that. 622 00:18:13.125 --> 00:18:15.375 I think they're all 64-bit now, but yeah, 623 00:18:15.375 --> 00:18:17.166 I know Raspberry Pi was like the 32-bit 624 00:18:17.166 --> 00:18:18.750 ARM that everyone had and 625 00:18:18.750 --> 00:18:20.250 was popular for quite a while. 626 00:18:21.333 --> 00:18:22.416 Yeah, it's funny. 627 00:18:22.833 --> 00:18:25.083 If you look at Windows or Mac, Mac is 628 00:18:25.083 --> 00:18:26.541 just Intel chip, right? 629 00:18:26.541 --> 00:18:29.041 It doesn't even think about 32-bits, and 630 00:18:29.041 --> 00:18:31.333 Windows, it's x64, x64, Arm64. 631 00:18:31.666 --> 00:18:33.250 They're using the terminology for each 632 00:18:33.250 --> 00:18:34.375 platform, which is nice. 633 00:18:35.416 --> 00:18:37.375 Anyway, that bundle structure that I just 634 00:18:37.375 --> 00:18:39.333 showed is the same on other platforms, 635 00:18:39.333 --> 00:18:41.416 because it's a very macOS thing to do to 636 00:18:41.416 --> 00:18:43.875 distribute an app as a folder and call it 637 00:18:43.875 --> 00:18:45.958 a bundle, and have it show differently in 638 00:18:45.958 --> 00:18:47.333 the finder, but actually it's just a 639 00:18:47.333 --> 00:18:48.625 folder if you look in the command line. 640 00:18:48.833 --> 00:18:50.833 But as far as I've been able to find, 641 00:18:50.833 --> 00:18:52.583 because I've been looking around the 642 00:18:52.583 --> 00:18:56.291 OFXRS directory, repository, even if I 643 00:18:56.291 --> 00:18:59.000 wasn't able to run it, I saw, I looked at 644 00:18:59.000 --> 00:19:01.375 how they were doing things, and they are 645 00:19:01.375 --> 00:19:04.625 making a content slash Linux dash x86 646 00:19:04.625 --> 00:19:06.666 dash 64 directory, which 647 00:19:06.666 --> 00:19:09.041 is, it's a Franken bundle. 648 00:19:09.416 --> 00:19:10.875 Like it's weird, because 649 00:19:10.875 --> 00:19:13.958 it's macOS semantics on Linux. 650 00:19:14.375 --> 00:19:15.041 I don't know. 651 00:19:15.250 --> 00:19:16.791 It's really weird to me. 652 00:19:16.791 --> 00:19:18.958 Anyway, NTSC RS does work. 653 00:19:18.958 --> 00:19:21.791 I was able to build it from source and 654 00:19:21.791 --> 00:19:23.250 import it in DaVinci 655 00:19:23.250 --> 00:19:26.250 Resolve, and it works everywhere. 656 00:19:26.250 --> 00:19:27.041 It works as an effect. 657 00:19:27.291 --> 00:19:28.250 You can drag onto a 658 00:19:28.250 --> 00:19:29.291 clip in the edit page. 659 00:19:29.291 --> 00:19:31.458 It works as an effect in the color page, 660 00:19:31.916 --> 00:19:34.583 and it works as a node in fusion, which 661 00:19:34.583 --> 00:19:36.000 is absolutely lovely. 662 00:19:36.000 --> 00:19:37.833 You can see on the slide is a me on a 663 00:19:37.833 --> 00:19:41.541 green screen with a NTSC effect on a 480p 664 00:19:41.541 --> 00:19:42.750 timeline, because that's 665 00:19:42.750 --> 00:19:43.666 what they recommend doing. 666 00:19:43.666 --> 00:19:44.833 Even though it can run at higher 667 00:19:44.833 --> 00:19:46.458 resolutions, it's not supposed to have 668 00:19:46.458 --> 00:19:47.916 more than 480 lines. 669 00:19:48.291 --> 00:19:50.083 You get sort of that like anachronism of 670 00:19:50.083 --> 00:19:52.875 like, here's my 4k image with scan lines, 671 00:19:53.125 --> 00:19:54.666 and you're like, that I don't think 672 00:19:54.666 --> 00:19:56.291 anyone was doing that. 673 00:19:56.750 --> 00:19:58.708 There might've been raw captures of that, 674 00:19:58.708 --> 00:20:00.750 but you certainly weren't broadcasting it 675 00:20:00.833 --> 00:20:02.375 more than 480 lines. 676 00:20:03.208 --> 00:20:04.416 It's got the vibe of it. 677 00:20:04.416 --> 00:20:07.458 It's very like found footage or like 678 00:20:07.458 --> 00:20:10.333 someone sent in a VHS of themselves in 679 00:20:10.333 --> 00:20:14.250 1990 to some technology show to it. 680 00:20:14.250 --> 00:20:15.291 You've definitely gotten 681 00:20:15.291 --> 00:20:16.416 that vibe out of the effect. 682 00:20:17.875 --> 00:20:19.833 And there's a ton of the parameters you 683 00:20:19.833 --> 00:20:21.708 can, a ton of controls you can change. 684 00:20:22.125 --> 00:20:24.833 I've put side by side the parameters as 685 00:20:24.833 --> 00:20:26.541 seen from the venture resolve and the 686 00:20:26.541 --> 00:20:28.583 parameters as seen from the 687 00:20:28.583 --> 00:20:30.250 standalone E-GUI interface. 688 00:20:30.750 --> 00:20:32.291 And the code base is just really 689 00:20:32.291 --> 00:20:34.291 well-designed because they have a single 690 00:20:34.291 --> 00:20:36.166 way to describe all the parameters and 691 00:20:36.166 --> 00:20:38.625 then it's mapped to whichever interface 692 00:20:39.375 --> 00:20:41.125 because they have essentially three 693 00:20:41.125 --> 00:20:42.083 different GUIs, right? 694 00:20:42.083 --> 00:20:43.541 They have egui and standalone, they 695 00:20:43.541 --> 00:20:45.250 have Premiere and After Effects, I think 696 00:20:45.250 --> 00:20:46.875 are roughly the same. 697 00:20:46.875 --> 00:20:47.958 I have a lot of commonalities. 698 00:20:48.416 --> 00:20:49.333 And then OpenFX. 699 00:20:49.916 --> 00:20:52.166 And OpenFX has different kinds of hosts 700 00:20:52.166 --> 00:20:53.458 like DaVinci Resolve 701 00:20:53.458 --> 00:20:56.750 or Natron or like any sort of VFX program 702 00:20:57.041 --> 00:20:58.125 that can load OpenFX 703 00:20:58.125 --> 00:20:59.291 plugins is called the host. 704 00:20:59.291 --> 00:21:00.833 It provides functions and provides these, 705 00:21:01.125 --> 00:21:03.333 the UI, the parameters like loading 706 00:21:03.333 --> 00:21:05.125 settings, saving settings, et cetera. 707 00:21:05.125 --> 00:21:08.166 And also you can see on every parameter 708 00:21:08.166 --> 00:21:09.208 here, there's a little 709 00:21:09.208 --> 00:21:11.166 diamond that's for keyframes. 710 00:21:11.583 --> 00:21:13.166 So OpenFX plugins don't 711 00:21:13.166 --> 00:21:14.125 need to know about keyframes. 712 00:21:14.250 --> 00:21:17.333 They just need to, for every render 713 00:21:17.333 --> 00:21:18.666 request, they just need to get the 714 00:21:18.666 --> 00:21:20.333 parameter for that time. 715 00:21:20.583 --> 00:21:22.375 And then the host is in charge of making 716 00:21:22.375 --> 00:21:24.125 the values change over time in the 717 00:21:24.125 --> 00:21:25.041 timeline and like 718 00:21:25.041 --> 00:21:26.500 interpolating things and stuff. 719 00:21:26.875 --> 00:21:27.291 Hello, Amos. 720 00:21:27.625 --> 00:21:29.166 Yes, as someone who does not edit a lot 721 00:21:29.166 --> 00:21:31.875 of video, what distinction does keyframes 722 00:21:31.875 --> 00:21:34.166 make when it comes to applying an effect? 723 00:21:34.166 --> 00:21:36.375 Is the effect not just a stream of images 724 00:21:36.375 --> 00:21:37.875 that it's modifying or is there 725 00:21:37.875 --> 00:21:38.875 intelligence going on? 726 00:21:38.916 --> 00:21:40.500 Sure, but you see intensity here. 727 00:21:40.750 --> 00:21:42.375 You can make it change over time. 728 00:21:43.000 --> 00:21:44.666 So keyframe is just like, it's this value 729 00:21:44.666 --> 00:21:45.875 at that frame and then you 730 00:21:45.875 --> 00:21:47.791 wait 30 frames and then change it. 731 00:21:47.791 --> 00:21:49.500 And then it changes linearly or like 732 00:21:49.500 --> 00:21:51.875 maybe there's a little tweening happening 733 00:21:52.166 --> 00:21:53.166 between the two values. 734 00:21:53.166 --> 00:21:54.375 It's just animation basically. 735 00:21:54.791 --> 00:21:55.208 Gotcha. 736 00:21:55.250 --> 00:21:56.916 So yeah, let's keep walking through like 737 00:21:56.916 --> 00:21:59.250 what is actually in an OpenFX plugin. 738 00:21:59.750 --> 00:22:01.250 There's a well-known struct. 739 00:22:01.500 --> 00:22:02.458 It's exactly what you 740 00:22:02.458 --> 00:22:03.958 would expect from a plugin API. 741 00:22:04.041 --> 00:22:06.000 Essentially there's well-known structs. 742 00:22:06.000 --> 00:22:08.833 You export two important functions. 743 00:22:09.125 --> 00:22:11.250 One is, ofxGetNumberOfPlugins which 744 00:22:11.250 --> 00:22:12.291 in this case returns one and 745 00:22:12.291 --> 00:22:13.208 it's not shown on the slide. 746 00:22:13.666 --> 00:22:15.875 And ofxGetPlugin, which returns a struct. 747 00:22:16.208 --> 00:22:18.416 What's interesting in the OpenFX API is 748 00:22:18.416 --> 00:22:19.375 that they use a lot of 749 00:22:19.375 --> 00:22:20.958 string constants everywhere. 750 00:22:21.416 --> 00:22:22.583 Sorry, the first thing in the struct that 751 00:22:22.583 --> 00:22:23.833 you return is just a magic string. 752 00:22:24.208 --> 00:22:27.375 It's a "kOfxImageEffectPluginApi" as a 753 00:22:27.375 --> 00:22:28.750 string and that's it, 754 00:22:28.750 --> 00:22:29.541 well, without the K. 755 00:22:29.541 --> 00:22:30.166 And then there's API 756 00:22:30.166 --> 00:22:32.041 version, which is good. 757 00:22:32.041 --> 00:22:33.791 That means they can actually evolve it. 758 00:22:33.791 --> 00:22:35.625 Well, it's set to one in this case, but 759 00:22:35.625 --> 00:22:36.916 they can just like add or-- 760 00:22:36.916 --> 00:22:38.375 Which means it's never changed. 761 00:22:38.875 --> 00:22:39.083 Yeah, 762 00:22:40.083 --> 00:22:42.208 but they could like change the struct. 763 00:22:42.458 --> 00:22:44.041 They could interpret the struct as having 764 00:22:44.041 --> 00:22:46.375 this header that never changes except for 765 00:22:46.375 --> 00:22:47.875 the API version field. 766 00:22:47.875 --> 00:22:49.916 And then based on if it's one or two or 767 00:22:49.916 --> 00:22:52.333 three, like read different fields after 768 00:22:52.333 --> 00:22:54.250 that, like sort of like a rust enum. 769 00:22:54.791 --> 00:22:56.208 Yeah, these are fun things for like 770 00:22:56.208 --> 00:22:58.333 binary compat, like, cause you run into 771 00:22:58.333 --> 00:23:00.416 it in protocols as well as like ABI 772 00:23:00.416 --> 00:23:02.708 things of, you need to just read the 773 00:23:02.708 --> 00:23:04.125 first two fields and then you promise 774 00:23:04.125 --> 00:23:05.250 that those never change. 775 00:23:05.750 --> 00:23:08.500 But that second API field, like you said, 776 00:23:08.500 --> 00:23:10.875 it acts like an enum where it's like, ah, 777 00:23:10.875 --> 00:23:12.750 do I actually know how to interpret the 778 00:23:12.750 --> 00:23:13.958 variant that it's claiming? 779 00:23:14.291 --> 00:23:16.083 Because if it says two and I have no idea 780 00:23:16.083 --> 00:23:17.833 what two means, I should stop reading 781 00:23:17.833 --> 00:23:18.791 because things are about 782 00:23:18.791 --> 00:23:20.125 to get very, very weird. 783 00:23:20.166 --> 00:23:21.000 Just throw an error. 784 00:23:21.083 --> 00:23:21.458 Yeah. 785 00:23:21.500 --> 00:23:22.291 Yeah, instead of 786 00:23:22.291 --> 00:23:23.083 getting memory corruption. 787 00:23:23.916 --> 00:23:25.750 Next up, we have main entry, which is 788 00:23:25.750 --> 00:23:26.958 directly one of the fields 789 00:23:26.958 --> 00:23:28.166 of that struct that we return. 790 00:23:28.583 --> 00:23:30.791 And main entry is kind of like the main, 791 00:23:30.791 --> 00:23:32.875 kind of like DLLs on Windows 792 00:23:32.875 --> 00:23:34.375 also have a main entry point. 793 00:23:34.750 --> 00:23:37.458 It's just funny to me because it gets a 794 00:23:37.458 --> 00:23:39.541 string, a C string and you compare it 795 00:23:39.541 --> 00:23:41.125 against a bunch of constants and then you 796 00:23:41.125 --> 00:23:42.666 know if the action you're supposed to do 797 00:23:42.666 --> 00:23:46.125 is load or describe or get regions of 798 00:23:46.125 --> 00:23:47.666 interest or whatever or render. 799 00:23:48.083 --> 00:23:51.000 So that means every frame you render, 800 00:23:51.208 --> 00:23:52.708 every tile you render, there's a 801 00:23:52.708 --> 00:23:54.625 multi-threading thing that 802 00:23:54.625 --> 00:23:56.000 you can ask the host to do. 803 00:23:56.000 --> 00:23:57.750 You can ask the host to like divide the 804 00:23:57.750 --> 00:23:59.500 image in different tiles and ask you to 805 00:23:59.500 --> 00:24:00.375 render tiles in 806 00:24:00.375 --> 00:24:01.625 different threads and whatnot. 807 00:24:01.625 --> 00:24:03.000 You're just a poor shared library. 808 00:24:03.083 --> 00:24:04.083 So your function can be 809 00:24:04.083 --> 00:24:04.875 called from different threads. 810 00:24:05.125 --> 00:24:07.083 The first thing you do is just compare 811 00:24:07.083 --> 00:24:09.750 against like do a chain of ifs. 812 00:24:09.750 --> 00:24:10.791 It's like, okay, it's not load. 813 00:24:10.791 --> 00:24:11.458 It's not described. 814 00:24:11.458 --> 00:24:12.708 It's not describing context, 815 00:24:12.708 --> 00:24:13.541 blah, blah, blah, blah, blah. 816 00:24:13.541 --> 00:24:14.791 Finally, I was like, oh, it is render. 817 00:24:15.208 --> 00:24:17.541 So either string comparisons are a lot 818 00:24:17.541 --> 00:24:21.583 faster than I imagine or that does not 819 00:24:21.583 --> 00:24:23.333 that worried about performance. 820 00:24:23.791 --> 00:24:25.000 I don't know, James, how 821 00:24:25.000 --> 00:24:26.000 do you feel about this? 822 00:24:26.083 --> 00:24:27.000 I mean, computers are fast. 823 00:24:27.458 --> 00:24:27.541 Like, yeah. 824 00:24:27.583 --> 00:24:27.791 True. 825 00:24:28.916 --> 00:24:30.458 I mean, a match, like it looks grosser 826 00:24:30.458 --> 00:24:32.000 than a match, but a match is essentially 827 00:24:32.000 --> 00:24:33.208 just doing the same thing. 828 00:24:33.500 --> 00:24:35.208 Like you can do much smarter string 829 00:24:35.208 --> 00:24:38.583 compares of like making either like a 830 00:24:38.583 --> 00:24:40.083 regex style automata 831 00:24:40.083 --> 00:24:40.791 or what are they called? 832 00:24:40.791 --> 00:24:41.791 They're not just finite statements. 833 00:24:42.041 --> 00:24:43.625 Sure, but why not just export constants 834 00:24:43.875 --> 00:24:45.166 from the library or something? 835 00:24:45.166 --> 00:24:48.458 Like define, I don't know, the U64 or 836 00:24:48.458 --> 00:24:51.041 whatever in file formats, like MP4, we 837 00:24:51.041 --> 00:24:52.916 have those atoms which 838 00:24:52.916 --> 00:24:55.250 are just 32 bits, like four. 839 00:24:55.541 --> 00:24:56.166 Yeah, four times 840 00:24:56.166 --> 00:24:57.875 eight is 32, just four characters. 841 00:24:58.291 --> 00:25:01.750 So you have like MOOV, you have MP4A, 842 00:25:01.750 --> 00:25:02.958 whatever, that's what all those things 843 00:25:02.958 --> 00:25:04.041 are four characters long. 844 00:25:04.041 --> 00:25:05.458 It's so that you get a nice 845 00:25:05.458 --> 00:25:07.250 32 bit constant out of it. 846 00:25:07.458 --> 00:25:11.000 I assume 32 str comps are way cheaper 847 00:25:11.000 --> 00:25:12.791 than whatever video effects you're about 848 00:25:12.791 --> 00:25:13.875 to be doing, honestly. 849 00:25:14.208 --> 00:25:15.833 So like order of magnitude, they might 850 00:25:15.833 --> 00:25:17.666 go, "Eh, it's so we can debug it easier." 851 00:25:17.666 --> 00:25:18.875 Cause then you could just print instead 852 00:25:18.875 --> 00:25:21.583 of having to do a lookup or whatever when 853 00:25:21.583 --> 00:25:22.708 you print logs of like, 854 00:25:22.708 --> 00:25:23.791 "We were told to do this." 855 00:25:24.125 --> 00:25:24.458 True. 856 00:25:24.500 --> 00:25:25.000 I don't know. 857 00:25:25.000 --> 00:25:25.833 I think you're right. 858 00:25:25.833 --> 00:25:27.375 It would definitely be much smaller, but 859 00:25:27.375 --> 00:25:28.750 I think probably one of those order of 860 00:25:28.750 --> 00:25:30.291 magnitude things where it goes, "Eh, 861 00:25:30.625 --> 00:25:31.625 whatever, we're about to do 862 00:25:31.625 --> 00:25:34.083 a bunch of DSP, so whatever." 863 00:25:34.541 --> 00:25:36.208 Well, what I also like is that in this, 864 00:25:36.208 --> 00:25:38.625 in NTSC-RS specifically, it calls set 865 00:25:38.625 --> 00:25:40.250 hook, set panic hook. 866 00:25:40.250 --> 00:25:40.875 And I have no idea 867 00:25:40.875 --> 00:25:41.916 what the cost of that is. 868 00:25:42.208 --> 00:25:43.375 Like I have not measured it. 869 00:25:43.375 --> 00:25:45.500 I've measured other parts because the 870 00:25:45.500 --> 00:25:47.625 plugin was slow at some point, I did some 871 00:25:47.625 --> 00:25:48.333 performance measurement. 872 00:25:48.625 --> 00:25:50.458 I did not measure how long it takes to 873 00:25:50.458 --> 00:25:51.625 set the panic hook, but 874 00:25:51.625 --> 00:25:53.416 I imagine, I don't know. 875 00:25:53.416 --> 00:25:55.208 I don't think a huge amount of time. 876 00:25:55.333 --> 00:25:56.500 No, but- Yeah, I have no idea 877 00:25:56.500 --> 00:25:57.625 either off the top of my head. 878 00:25:57.625 --> 00:25:58.958 This is running in std, which 879 00:25:58.958 --> 00:26:00.375 I don't do most of the time. 880 00:26:00.791 --> 00:26:01.750 Or at least I don't think too 881 00:26:01.750 --> 00:26:02.958 hard when I'm running in std. 882 00:26:02.958 --> 00:26:04.083 Like I think hard when I'm running in 883 00:26:04.083 --> 00:26:05.416 bare metal, but yeah. 884 00:26:05.875 --> 00:26:07.500 So next up, if the action that we're 885 00:26:07.500 --> 00:26:10.125 supposed to do is a render, then we get 886 00:26:10.125 --> 00:26:13.833 the address of the functions from the 887 00:26:13.833 --> 00:26:16.000 host that we're supposed to call, like 888 00:26:16.000 --> 00:26:17.833 propGetString or 889 00:26:17.833 --> 00:26:19.250 propGetDouble, et cetera. 890 00:26:19.250 --> 00:26:21.916 Then we call that and we get all the 891 00:26:21.916 --> 00:26:22.833 properties that we want. 892 00:26:22.833 --> 00:26:25.333 Again, mostly with string constants, like 893 00:26:25.333 --> 00:26:26.208 the propGetDouble, 894 00:26:26.541 --> 00:26:28.333 kOfxPropTime.as_ptr() 895 00:26:28.333 --> 00:26:29.375 That's just the string. 896 00:26:29.583 --> 00:26:30.750 That's just "OfxPropTime". 897 00:26:32.208 --> 00:26:34.250 And also we're getting it as a double, 898 00:26:34.250 --> 00:26:37.000 which is an F64, but it's actually just 899 00:26:37.000 --> 00:26:37.875 in DaVinci Resolve, 900 00:26:37.875 --> 00:26:39.000 that's the frame number. 901 00:26:39.833 --> 00:26:41.083 So it's fully an integer. 902 00:26:41.416 --> 00:26:42.958 speed, summer games done quick was just 903 00:26:42.958 --> 00:26:44.166 happened, which means everyone was 904 00:26:44.166 --> 00:26:46.416 talking about basically all like speed 905 00:26:46.416 --> 00:26:47.208 running and glitching 906 00:26:47.208 --> 00:26:48.083 and stuff like that. 907 00:26:48.083 --> 00:26:50.083 And there's a whole thing of like, what 908 00:26:50.083 --> 00:26:52.625 goes wrong when you use a floating point 909 00:26:52.625 --> 00:26:54.166 number for your frame counter. 910 00:26:54.458 --> 00:26:56.000 And it was basically like 16 million is 911 00:26:56.000 --> 00:26:59.208 where just add one breaks down because it 912 00:26:59.208 --> 00:27:01.125 rounds down instead of up, which means 913 00:27:01.125 --> 00:27:03.833 you just hard lock at 16 million frames. 914 00:27:03.875 --> 00:27:07.958 So 16 millions at 25 FPS. 915 00:27:08.833 --> 00:27:10.583 Oh, it's like months. 916 00:27:10.916 --> 00:27:12.791 Like this was a Mario 64 glitch that was 917 00:27:12.791 --> 00:27:15.000 like, look, this portal stops rippling 918 00:27:15.000 --> 00:27:17.500 after four and a half months, but this 919 00:27:17.500 --> 00:27:19.583 one stops after a month and a half 920 00:27:19.833 --> 00:27:20.875 because it increments by 921 00:27:20.875 --> 00:27:22.291 two or something like that. 922 00:27:22.291 --> 00:27:24.625 It reminds me of the Windows 95 thing 923 00:27:24.625 --> 00:27:26.291 where if it stays up 924 00:27:26.375 --> 00:27:28.750 for 91 days or something, 925 00:27:29.458 --> 00:27:30.375 didn't do this. 926 00:27:30.625 --> 00:27:32.166 You get this in planes and stuff. 927 00:27:32.166 --> 00:27:32.708 When you have- 928 00:27:32.708 --> 00:27:33.583 It overflows. 929 00:27:33.625 --> 00:27:34.750 It's either a millisecond 930 00:27:34.750 --> 00:27:36.250 or a 10th of a millisecond. 931 00:27:36.500 --> 00:27:37.791 Whenever you see something that says it 932 00:27:37.791 --> 00:27:40.291 needs to be restarted every 50 days, or 933 00:27:40.291 --> 00:27:42.708 it crashes every 50 days, that's like, 934 00:27:43.208 --> 00:27:45.500 yeah, it's either every millisecond or a 935 00:27:45.500 --> 00:27:48.208 10th of a millisecond in a U32 counter. 936 00:27:48.750 --> 00:27:50.166 Or like, yeah, you start 937 00:27:50.166 --> 00:27:51.208 to notice those numbers. 938 00:27:51.541 --> 00:27:53.083 Like this is one millisecond, 10 939 00:27:53.083 --> 00:27:55.083 millisecond, a hundred millisecond, one 940 00:27:55.083 --> 00:27:56.875 microsecond of like those times where 941 00:27:56.875 --> 00:27:58.791 it's like, yeah, it just crashes if you 942 00:27:58.791 --> 00:28:00.208 run that long and it means that they 943 00:28:00.208 --> 00:28:01.583 didn't handle overflows right. 944 00:28:01.916 --> 00:28:04.375 Well, to be fair, it can be hard to 945 00:28:04.375 --> 00:28:06.250 handle that properly. 16 million frames is 946 00:28:06.250 --> 00:28:08.916 25, according to o3, which 947 00:28:08.916 --> 00:28:11.000 I trust with reservations, 948 00:28:12.333 --> 00:28:14.583 it's seven days, nine hours, 46 minutes 949 00:28:14.583 --> 00:28:15.625 and 40 seconds. 16 950 00:28:15.625 --> 00:28:17.958 So if you make a student film, that's 951 00:28:17.958 --> 00:28:19.291 just really boring in 952 00:28:19.291 --> 00:28:19.833 DaVinci Resolve. 953 00:28:20.208 --> 00:28:21.875 You might get into issues at some point. 954 00:28:22.291 --> 00:28:24.208 So you get all your, you get the 955 00:28:24.208 --> 00:28:27.250 addresses of the methods you need to call 956 00:28:27.250 --> 00:28:30.333 from the host, and then you call those 957 00:28:30.333 --> 00:28:32.166 and you get all the properties, all the 958 00:28:32.166 --> 00:28:34.833 parameters that I've shown earlier, like 959 00:28:34.833 --> 00:28:37.041 intensity, frequency, detail, snow, in 960 00:28:37.041 --> 00:28:38.458 the case of NTSC RS, but 961 00:28:38.458 --> 00:28:39.333 it could be anything, right? 962 00:28:39.333 --> 00:28:40.291 You can have anything in 963 00:28:40.291 --> 00:28:41.541 the UI, it's pretty flexible. 964 00:28:41.875 --> 00:28:45.708 You also get the input signal, the output 965 00:28:45.708 --> 00:28:47.125 signal, output buffer, where you're 966 00:28:47.125 --> 00:28:48.750 supposed to write the thing, the result 967 00:28:48.750 --> 00:28:49.916 of your effect being applied. 968 00:28:50.166 --> 00:28:52.583 The host tells you which part of the 969 00:28:52.583 --> 00:28:53.958 output buffer you're supposed to touch. 970 00:28:54.375 --> 00:28:57.166 It tells you the buffer format, if it's 971 00:28:57.166 --> 00:29:00.708 RGB, RGBA, if it's eight, 16, 32 bits, if 972 00:29:00.708 --> 00:29:02.333 it's floating point integer. 973 00:29:02.708 --> 00:29:05.666 I was curious what code path actually got 974 00:29:05.666 --> 00:29:07.458 used because NTSC RS 975 00:29:07.458 --> 00:29:08.833 handles all possible combinations. 976 00:29:09.416 --> 00:29:10.750 So how is developing 977 00:29:10.750 --> 00:29:11.708 an open effects plugin? 978 00:29:11.958 --> 00:29:14.083 I wanted to find out what 979 00:29:14.083 --> 00:29:16.125 was the actual buffer format. 980 00:29:16.125 --> 00:29:16.708 I wasn't planning on 981 00:29:16.708 --> 00:29:17.541 supporting everything. 982 00:29:17.875 --> 00:29:20.291 This is research level, we're just trying 983 00:29:20.291 --> 00:29:21.750 to get it to work for one specific 984 00:29:21.750 --> 00:29:23.041 configuration, like only 985 00:29:23.041 --> 00:29:26.000 on Mac OS, only on one host. 986 00:29:26.625 --> 00:29:27.708 And the first thing that 987 00:29:27.708 --> 00:29:29.041 hit me is how do you log? 988 00:29:29.375 --> 00:29:30.166 James, do you have any idea? 989 00:29:30.416 --> 00:29:31.166 How do you log? 990 00:29:31.208 --> 00:29:31.708 How do you log? 991 00:29:31.708 --> 00:29:32.583 How do you print things? 992 00:29:32.916 --> 00:29:35.083 Okay, so if I had to guess, there's... 993 00:29:35.083 --> 00:29:36.208 You send a 994 00:29:36.208 --> 00:29:37.791 postcard-rpc messag—no. 995 00:29:38.833 --> 00:29:39.750 It's probably not far off. 996 00:29:39.750 --> 00:29:40.916 I mean, what I would assume is in the 997 00:29:40.916 --> 00:29:42.750 plugin, you get passed in some vtabl 998 00:29:42.750 --> 00:29:43.625 that has methods that 999 00:29:43.625 --> 00:29:44.375 you're allowed to call. 1000 00:29:45.041 --> 00:29:45.916 Or there's some like 1001 00:29:45.916 --> 00:29:47.458 global or something like that. 1002 00:29:47.458 --> 00:29:47.833 But no. 1003 00:29:47.875 --> 00:29:48.375 Yes, but no. 1004 00:29:48.416 --> 00:29:50.125 There's absolutely nothing standard. 1005 00:29:50.541 --> 00:29:51.750 Oh, love it, love it. 1006 00:29:51.750 --> 00:29:52.416 There's no OpenFX 1007 00:29:52.416 --> 00:29:53.375 blessed way to do logs. 1008 00:29:53.833 --> 00:29:56.333 You can write to standard error, but then 1009 00:29:56.333 --> 00:29:59.208 you need to start your host from the 1010 00:29:59.208 --> 00:30:00.833 terminal, which is annoying. 1011 00:30:01.625 --> 00:30:02.291 Resolver has its own 1012 00:30:02.291 --> 00:30:03.416 log, which you can't touch. 1013 00:30:03.416 --> 00:30:04.833 I'm assuming you can write it from fuses 1014 00:30:05.166 --> 00:30:07.750 and they also have their own SDK to make 1015 00:30:07.750 --> 00:30:09.125 plugins that are not open effects. 1016 00:30:09.541 --> 00:30:11.125 Are you gonna tell me that you spawned a 1017 00:30:11.125 --> 00:30:12.833 Lua interpreter just to print messages 1018 00:30:12.833 --> 00:30:14.791 using some kind of Lua trampoline or 1019 00:30:14.791 --> 00:30:15.375 something like that? 1020 00:30:15.416 --> 00:30:16.666 No, I only had a day, James. 1021 00:30:16.666 --> 00:30:17.250 I didn't do that. 1022 00:30:17.250 --> 00:30:18.833 No, no, no, you're seeing solution in 1023 00:30:18.833 --> 00:30:20.416 front of you, which I just opened the 1024 00:30:20.416 --> 00:30:21.958 file and just appended to it. 1025 00:30:22.333 --> 00:30:24.583 Yeah, okay, yeah, just no sandboxing, 1026 00:30:24.583 --> 00:30:25.291 just straight to temp. 1027 00:30:25.708 --> 00:30:26.958 Yeah, of course. 1028 00:30:29.791 --> 00:30:32.750 Sandboxing, James, your concerns are very 1029 00:30:32.750 --> 00:30:33.875 far from the concerns 1030 00:30:33.875 --> 00:30:35.333 of VFX people, I think. 1031 00:30:35.375 --> 00:30:36.625 Oh, it's fair, it's fair, yeah. 1032 00:30:37.583 --> 00:30:38.416 Another question, how do 1033 00:30:38.416 --> 00:30:39.750 you reload your plugin? 1034 00:30:40.333 --> 00:30:40.750 Any ideas? 1035 00:30:41.625 --> 00:30:41.958 You panic. 1036 00:30:42.458 --> 00:30:44.625 No, if you panic, I'm pretty sure it 1037 00:30:44.625 --> 00:30:45.291 takes down the host. 1038 00:30:45.375 --> 00:30:45.708 Okay. 1039 00:30:45.708 --> 00:30:46.375 Well, yeah. 1040 00:30:46.416 --> 00:30:47.958 Oh, great, yeah, that's great. 1041 00:30:48.250 --> 00:30:49.166 Write no sandboxing. 1042 00:30:49.583 --> 00:30:50.333 Yeah, this is not 1043 00:30:50.333 --> 00:30:52.208 WebAssembly, this is not, you know. 1044 00:30:52.291 --> 00:30:54.708 How do you hotload or reload your bundle? 1045 00:30:55.041 --> 00:30:56.625 Exactly, well, you don't, 1046 00:30:56.625 --> 00:30:57.125 you can't. 1047 00:30:57.916 --> 00:30:59.250 I'm showing you now the video plugin 1048 00:30:59.250 --> 00:31:02.166 settings pane and you can do, "pain" is 1049 00:31:02.166 --> 00:31:04.208 right, and you can disable all, enable 1050 00:31:04.208 --> 00:31:07.000 all, but all it does is just show you the 1051 00:31:07.000 --> 00:31:08.083 plugin and the list of 1052 00:31:08.083 --> 00:31:09.416 things you can use or not. 1053 00:31:09.791 --> 00:31:12.000 But once it's loaded, it's loaded, right? 1054 00:31:12.375 --> 00:31:14.875 "dlclose" is a lie in 1055 00:31:14.875 --> 00:31:15.916 pretty much all platforms. 1056 00:31:15.916 --> 00:31:17.250 It always leaks things. 1057 00:31:17.541 --> 00:31:19.500 There's no cleanly unloading something. 1058 00:31:19.750 --> 00:31:21.541 In Rust, if you load a library, it 1059 00:31:21.541 --> 00:31:23.791 returns to static lifetime and for good 1060 00:31:23.791 --> 00:31:25.416 reasons because you really shouldn't 1061 00:31:25.416 --> 00:31:26.875 unload things that you've loaded. 1062 00:31:26.875 --> 00:31:29.166 Even if you could like cleanly do it, 1063 00:31:29.166 --> 00:31:30.958 apparently in C++ you can cleanly do it. 1064 00:31:31.375 --> 00:31:32.833 Even if you could do all that and you're 1065 00:31:32.833 --> 00:31:36.291 initializing metal kernels and like 1066 00:31:36.291 --> 00:31:39.000 shaders and whatever and those can't 1067 00:31:39.000 --> 00:31:41.625 necessarily be torn down pretty like, I 1068 00:31:41.625 --> 00:31:43.500 don't know, I don't know the details, but 1069 00:31:43.500 --> 00:31:44.333 essentially you cannot. 1070 00:31:44.750 --> 00:31:47.750 So the way to reload it is to close the 1071 00:31:47.750 --> 00:31:49.458 Vintra Resolve and then reopen the Vintra 1072 00:31:49.458 --> 00:31:50.833 Resolve and then reload the project. 1073 00:31:51.083 --> 00:31:52.125 Do you have any concept 1074 00:31:52.125 --> 00:31:53.208 of how long that takes? 1075 00:31:54.000 --> 00:31:55.291 Uh, like? 1076 00:31:55.583 --> 00:31:56.166 I've timed it. 1077 00:31:56.333 --> 00:31:57.291 How long to shut it down? 1078 00:31:57.333 --> 00:31:59.541 Okay, I know vaguely what it is for the 1079 00:31:59.541 --> 00:32:01.291 self-directed research, but we do no 1080 00:32:01.291 --> 00:32:03.083 video effects or any, well, I mean like 1081 00:32:03.083 --> 00:32:05.791 we have slides, I'm sure, minutes? 1082 00:32:06.458 --> 00:32:08.916 No, no, thankfully, no, 1083 00:32:08.916 --> 00:32:10.791 I have a very nice M4. 1084 00:32:11.208 --> 00:32:12.416 Wait, is it taking minutes on 1085 00:32:12.416 --> 00:32:14.000 your end to open the project? 1086 00:32:14.291 --> 00:32:16.750 No, it takes like 15 or 20 seconds for 1087 00:32:16.750 --> 00:32:17.958 this, but we also 1088 00:32:17.958 --> 00:32:18.833 aren't loading any video. 1089 00:32:19.041 --> 00:32:19.750 Yeah, exactly, yeah, yeah, 1090 00:32:19.750 --> 00:32:20.791 that's more in the ballpark. 1091 00:32:21.250 --> 00:32:22.833 What gets me is that it takes 10 seconds 1092 00:32:22.833 --> 00:32:24.333 to shut down before you 1093 00:32:24.333 --> 00:32:25.291 can even start it up again. 1094 00:32:25.750 --> 00:32:26.250 That's annoying. 1095 00:32:26.625 --> 00:32:28.916 You know someone just has a sleep that 1096 00:32:28.916 --> 00:32:30.208 they're like waiting for something to do 1097 00:32:30.250 --> 00:32:31.708 and they're like, whatever, it's a bunch 1098 00:32:31.708 --> 00:32:34.208 of sleep(1) or sleep(3) in there and 1099 00:32:34.208 --> 00:32:35.125 it just, it becomes 1100 00:32:35.125 --> 00:32:36.791 serialized and it takes 10 seconds. 1101 00:32:37.125 --> 00:32:39.208 I think it's all the shaders and stuff, 1102 00:32:39.208 --> 00:32:40.750 whatever, being unloaded or something. 1103 00:32:41.000 --> 00:32:43.250 Actually, oh, Jesus, 1104 00:32:43.458 --> 00:32:44.458 that's another nerd snipe. 1105 00:32:44.708 --> 00:32:46.625 But Instruments could probably tell 1106 00:32:46.625 --> 00:32:48.416 us what it's spending its time on. 1107 00:32:48.666 --> 00:32:51.166 Come back next season for debugging the 1108 00:32:51.166 --> 00:32:52.625 plug-in interface of this, but yeah. 1109 00:32:53.708 --> 00:32:55.625 No, this is the end of the arc. 1110 00:32:57.333 --> 00:32:58.750 That is one of those fun things of the 1111 00:32:58.750 --> 00:33:01.750 thundering herd of destructors is a not 1112 00:33:01.750 --> 00:33:04.000 trivial performance problem in a lot of 1113 00:33:04.000 --> 00:33:06.000 places of just, oh, we're gonna run all 1114 00:33:06.000 --> 00:33:08.041 the destructors now and it turns out that 1115 00:33:08.041 --> 00:33:10.125 takes a long time, but actually just 1116 00:33:10.125 --> 00:33:11.458 killing the program would have achieved 1117 00:33:11.458 --> 00:33:14.500 the same thing 99% of the time, except 1118 00:33:14.500 --> 00:33:16.291 for now it takes us 30 1119 00:33:16.291 --> 00:33:17.875 seconds to close the program. 1120 00:33:18.458 --> 00:33:20.875 So because it takes a long time to shut 1121 00:33:20.875 --> 00:33:22.333 down DaVinci Resolve and open it up again 1122 00:33:22.583 --> 00:33:25.625 what other people do is that they use 1123 00:33:25.625 --> 00:33:28.166 another host like Natron, I don't know 1124 00:33:28.166 --> 00:33:29.083 how to say it in English. 1125 00:33:29.083 --> 00:33:30.625 I know it's a French software. 1126 00:33:30.875 --> 00:33:33.500 It's scriptable with Python and you just 1127 00:33:33.500 --> 00:33:36.625 write a little script that sets up a node 1128 00:33:36.625 --> 00:33:37.625 graph and renders 1129 00:33:37.625 --> 00:33:39.916 everything and saves an image. 1130 00:33:40.208 --> 00:33:40.833 Now, Amos, I can 1131 00:33:40.833 --> 00:33:42.458 excuse a Lua, but a Python? 1132 00:33:45.000 --> 00:33:47.083 I told you, it's also doing 1133 00:33:47.083 --> 00:33:49.125 Python in DaVinci Resolve. 1134 00:33:49.291 --> 00:33:50.666 Sure, I remember what episode it was. 1135 00:33:50.666 --> 00:33:52.291 There was definitely some episode where I 1136 00:33:52.291 --> 00:33:54.333 said something about Python and Amos was 1137 00:33:54.333 --> 00:33:55.708 like, I can excuse a 1138 00:33:55.708 --> 00:33:56.666 Forth, but a Python? 1139 00:33:57.333 --> 00:33:57.833 Oh, I think it was 1140 00:33:57.833 --> 00:33:58.875 talking about bootstrapping. 1141 00:33:59.416 --> 00:34:03.500 But there's no aarch64 version of Natron or 1142 00:34:03.500 --> 00:34:04.541 Natron or something. 1143 00:34:04.833 --> 00:34:06.250 When I launch it on my computer, it just 1144 00:34:06.250 --> 00:34:07.291 immediately crashes. 1145 00:34:07.625 --> 00:34:08.291 Well, not immediately, 1146 00:34:08.625 --> 00:34:09.500 that's what's puzzling. 1147 00:34:09.750 --> 00:34:11.458 Like most things immediately crash. 1148 00:34:11.458 --> 00:34:13.750 It opens, it shows a little window, waits 1149 00:34:13.750 --> 00:34:15.958 a few seconds, it's loading more stuff 1150 00:34:15.958 --> 00:34:17.000 and then it crashes. 1151 00:34:17.375 --> 00:34:20.958 At that point, we have talked about 1152 00:34:20.958 --> 00:34:22.708 OpenFX, but we haven't really talked 1153 00:34:22.708 --> 00:34:23.750 about how do we actually 1154 00:34:23.750 --> 00:34:24.916 do the rendering, right? 1155 00:34:24.958 --> 00:34:26.416 We wanted to improve on the text 1156 00:34:26.416 --> 00:34:27.416 rendering with Blackmagic Fusion. 1157 00:34:28.000 --> 00:34:29.208 How do we actually do that in Rust? 1158 00:34:29.500 --> 00:34:30.333 James, I don't know if you're familiar 1159 00:34:30.541 --> 00:34:31.166 with the font 1160 00:34:31.166 --> 00:34:33.666 rendering ecosystem in Rust. 1161 00:34:34.166 --> 00:34:35.916 No, no is my answer. 1162 00:34:35.916 --> 00:34:37.375 I know the people who are doing it a lot 1163 00:34:37.375 --> 00:34:38.291 of the times, but I am 1164 00:34:38.291 --> 00:34:39.208 not familiar with it. 1165 00:34:39.208 --> 00:34:39.500 You do? 1166 00:34:39.958 --> 00:34:40.416 Well, okay. 1167 00:34:41.541 --> 00:34:41.750 (Laughs) 1168 00:34:41.750 --> 00:34:42.583 That was a hell of a reaction. 1169 00:34:42.833 --> 00:34:44.500 Okay, I wanted, what I was thinking when 1170 00:34:44.500 --> 00:34:46.208 I blurted out was okay, was like, you 1171 00:34:46.208 --> 00:34:46.916 choose to thank them, 1172 00:34:46.916 --> 00:34:47.833 they're doing great work. 1173 00:34:48.041 --> 00:34:49.250 But what came out was okay. 1174 00:34:49.541 --> 00:34:49.833 Oh. 1175 00:34:49.833 --> 00:34:50.208 Oh. 1176 00:34:50.250 --> 00:34:51.125 Well, I know some of the people. 1177 00:34:51.583 --> 00:34:53.416 I don't recognize Mooman219. 1178 00:34:53.791 --> 00:34:55.333 If it wasn't Rafe, then I probably 1179 00:34:55.333 --> 00:34:57.291 wouldn't have recognized who it was. 1180 00:34:57.583 --> 00:34:58.916 The one that I found that is pretty 1181 00:34:58.916 --> 00:35:01.958 simple and that kind of replaces rusttype 1182 00:35:01.958 --> 00:35:05.541 type and ab-glyph and glyphbrush and 1183 00:35:05.541 --> 00:35:08.250 glyphbrush-layout, rusttype being like, 1184 00:35:08.250 --> 00:35:09.875 I think an early port 1185 00:35:10.291 --> 00:35:12.083 of free type to Rust. 1186 00:35:12.500 --> 00:35:13.416 Pure Rust alternative. 1187 00:35:14.166 --> 00:35:15.416 I'm not actually sure if it's a port. 1188 00:35:15.666 --> 00:35:17.083 I'm just excited that it's no_std 1189 00:35:17.083 --> 00:35:19.000 because now I need to, I want to go mess 1190 00:35:19.000 --> 00:35:20.583 with it and see if I can use that for 1191 00:35:20.583 --> 00:35:21.958 font rendering because that's a fun thing 1192 00:35:21.958 --> 00:35:23.583 in embedded where a lot of the times you 1193 00:35:23.583 --> 00:35:24.750 pre-bake your fonts. 1194 00:35:25.208 --> 00:35:27.041 Like you rasterize them to specific font 1195 00:35:27.041 --> 00:35:28.208 sizes and then you just-- 1196 00:35:28.250 --> 00:35:28.958 Yeah, a lot of this 1197 00:35:28.958 --> 00:35:30.166 stuff is actually no_std 1198 00:35:30.416 --> 00:35:32.333 Take the glyphs and stamp them in. 1199 00:35:32.333 --> 00:35:32.583 Yeah. 1200 00:35:32.625 --> 00:35:32.833 Nice. 1201 00:35:33.250 --> 00:35:35.000 This is how far I got with my research. 1202 00:35:35.291 --> 00:35:37.000 I integrated fontdue, like I kind of 1203 00:35:37.000 --> 00:35:40.416 vampirized the NTSC-RS plugin and I got it 1204 00:35:40.416 --> 00:35:42.416 to render the number of 1205 00:35:42.416 --> 00:35:43.916 the frame in the output. 1206 00:35:44.250 --> 00:35:46.250 At first upside down and 1207 00:35:46.250 --> 00:35:47.750 then finally right side up. 1208 00:35:47.833 --> 00:35:49.708 Is this one of those like is zero top 1209 00:35:49.708 --> 00:35:52.291 left or bottom left or bottom right? 1210 00:35:52.541 --> 00:35:53.791 Who is it that has the excellent someone 1211 00:35:54.041 --> 00:35:55.833 we both follow on Twitter and then 1212 00:35:55.833 --> 00:35:58.458 BlueSky has all of the graphic, are all like 1213 00:35:58.458 --> 00:35:59.916 the game engines and stuff like that? 1214 00:36:00.166 --> 00:36:01.416 I think it's Freya Holmer. 1215 00:36:01.916 --> 00:36:02.708 Oh yeah, exactly. 1216 00:36:02.916 --> 00:36:05.125 Exactly, it's Freya of like, what is X, Y 1217 00:36:05.125 --> 00:36:07.041 and Z positive, negative and where does 1218 00:36:07.041 --> 00:36:08.791 the zero index start on all 1219 00:36:08.791 --> 00:36:10.625 of those on 2D and 3D modes? 1220 00:36:10.625 --> 00:36:12.125 Cause everything's different. 1221 00:36:12.541 --> 00:36:14.250 But that's the fun thing is if there's 1222 00:36:14.250 --> 00:36:16.625 several ways to do a thing, if there's 1223 00:36:16.625 --> 00:36:18.666 several possible conventions, all of them 1224 00:36:18.666 --> 00:36:19.583 are in use somewhere. 1225 00:36:20.708 --> 00:36:21.958 But Unreal is actually 1226 00:36:21.958 --> 00:36:23.250 changing that I think. 1227 00:36:23.333 --> 00:36:23.791 Oh really? 1228 00:36:24.041 --> 00:36:24.458 Okay. 1229 00:36:24.541 --> 00:36:25.083 Yeah, they're 1230 00:36:25.083 --> 00:36:27.250 switching to something else. 1231 00:36:27.666 --> 00:36:28.291 I don't know, they're 1232 00:36:28.291 --> 00:36:29.500 joining it or whatever. 1233 00:36:29.958 --> 00:36:30.833 I saw the announcement, 1234 00:36:30.833 --> 00:36:31.750 I was like, what, why? 1235 00:36:32.125 --> 00:36:33.333 It's gonna break everything. 1236 00:36:33.875 --> 00:36:34.791 I'm pretty happy with that 1237 00:36:34.791 --> 00:36:36.708 and it runs in almost real time. 1238 00:36:36.916 --> 00:36:38.500 Apparently getting all the parameters 1239 00:36:38.500 --> 00:36:40.583 that I wasn't using was 1240 00:36:40.583 --> 00:36:42.375 what was taking the most time. 1241 00:36:42.458 --> 00:36:45.625 And I got it down to like for a 4K frame 1242 00:36:45.625 --> 00:36:47.833 rendering between three and four 1243 00:36:47.833 --> 00:36:48.750 milliseconds, which I 1244 00:36:48.750 --> 00:36:49.916 thought was pretty nice. 1245 00:36:50.166 --> 00:36:53.333 I thought we had like a, for 60 FPS, you 1246 00:36:53.333 --> 00:36:56.000 get a 16 millisecond budget. 1247 00:36:56.000 --> 00:36:58.416 Yeah, 16.6 milliseconds of frame, yeah. 1248 00:36:58.666 --> 00:37:00.750 Exactly and I only need to do 25 FPS. 1249 00:37:01.166 --> 00:37:03.083 So I figured we had plenty of time but 1250 00:37:03.083 --> 00:37:04.208 it's not running smoothly. 1251 00:37:04.458 --> 00:37:05.166 You know what that means? 1252 00:37:05.416 --> 00:37:06.791 You know what that means, Amos, right? 1253 00:37:06.791 --> 00:37:07.375 What does that mean? 1254 00:37:07.708 --> 00:37:09.041 You were using the NTSC 1255 00:37:09.041 --> 00:37:10.791 plugin at a PAL framerate. 1256 00:37:13.416 --> 00:37:15.250 Because NTSC is 30 or 1257 00:37:15.250 --> 00:37:18.833 59.97 and PAL is 25 or 50. 1258 00:37:18.875 --> 00:37:20.208 But my footage is, I 1259 00:37:20.208 --> 00:37:21.166 live in Europe, okay? 1260 00:37:21.458 --> 00:37:21.916 Leave me alone. 1261 00:37:22.958 --> 00:37:23.291 But yeah. 1262 00:37:23.833 --> 00:37:25.541 Yeah, it's like the slow-mo guys episode 1263 00:37:25.541 --> 00:37:27.125 where they explain why they want the 1264 00:37:27.125 --> 00:37:29.166 lower frame rate and it's just why more 1265 00:37:29.166 --> 00:37:30.583 frames doesn't actually help. 1266 00:37:30.916 --> 00:37:33.583 Yeah, for some reason, even with it 1267 00:37:33.583 --> 00:37:35.125 taking three or four milliseconds by 1268 00:37:35.125 --> 00:37:37.750 frame, it's still not running smoothly. 1269 00:37:38.291 --> 00:37:39.541 Like playback is not smooth. 1270 00:37:39.541 --> 00:37:41.458 Fortunately, you have a playback 1271 00:37:41.458 --> 00:37:43.000 resolution kind of fixed. 1272 00:37:43.000 --> 00:37:44.791 You can run at half resolution while you 1273 00:37:44.791 --> 00:37:46.291 do the editing and then while 1274 00:37:46.291 --> 00:37:47.833 you export it full resolution. 1275 00:37:48.208 --> 00:37:50.583 But it's still kind of not great because 1276 00:37:50.583 --> 00:37:53.000 this is just one line of very basic text. 1277 00:37:53.458 --> 00:37:55.166 This is why I was wondering about this, 1278 00:37:55.583 --> 00:37:57.666 setting the panic hook because maybe 1279 00:37:57.666 --> 00:37:58.541 that's taking a few 1280 00:37:58.541 --> 00:38:00.041 milliseconds here and there, you know? 1281 00:38:00.375 --> 00:38:01.708 Probably not much but... 1282 00:38:01.750 --> 00:38:02.666 Yeah, I was just saying, if you don't 1283 00:38:02.666 --> 00:38:05.166 have logging, then doing timestamps of 1284 00:38:05.166 --> 00:38:06.791 things is gonna be egregious. 1285 00:38:07.375 --> 00:38:09.416 And then, like you said, trying to do 1286 00:38:09.416 --> 00:38:11.958 profiling for a DLL that you didn't 1287 00:38:11.958 --> 00:38:13.541 start, that someone else is starting 1288 00:38:13.541 --> 00:38:16.125 becomes dot, dot, dot fun. 1289 00:38:16.583 --> 00:38:18.875 Especially if the Egooey, like if that 1290 00:38:18.875 --> 00:38:21.708 raw binary setup does things that are 1291 00:38:21.708 --> 00:38:23.666 just intrinsically different than the 1292 00:38:23.666 --> 00:38:26.083 inside of DaVinci Resolve kind of things 1293 00:38:26.291 --> 00:38:28.250 or inside of Fusion or whatever. 1294 00:38:29.041 --> 00:38:30.250 Huh, that sounds fun. 1295 00:38:30.250 --> 00:38:31.875 Also, the other issue is that this 1296 00:38:31.875 --> 00:38:36.125 doesn't actually improve on the renderer 1297 00:38:36.125 --> 00:38:37.333 from Fusion because it 1298 00:38:37.333 --> 00:38:38.958 does not support color emojis. 1299 00:38:39.291 --> 00:38:40.250 It just supports like 1300 00:38:40.250 --> 00:38:41.291 the basic free type stuff. 1301 00:38:41.541 --> 00:38:42.541 Look basic, I don't know. 1302 00:38:43.000 --> 00:38:45.125 Look at me dismissing their hard work. 1303 00:38:45.125 --> 00:38:46.541 It's still like an amazing feat. 1304 00:38:46.875 --> 00:38:48.541 It's only basic, yeah. 1305 00:38:48.625 --> 00:38:50.291 But it supports a ton of 1306 00:38:50.291 --> 00:38:51.958 stuff but not color emojis. 1307 00:38:52.375 --> 00:38:55.333 Swash supports more things as far as I 1308 00:38:55.333 --> 00:38:58.375 can tell but it doesn't do layouting. 1309 00:38:58.375 --> 00:38:59.833 If you wanna do layouting, you need 1310 00:38:59.833 --> 00:39:01.458 something like cosmic text. 1311 00:39:02.000 --> 00:39:05.000 And cosmic text actually works really 1312 00:39:05.000 --> 00:39:07.291 nice but I tried the demo 1313 00:39:07.291 --> 00:39:08.916 and it does render color emoji. 1314 00:39:08.916 --> 00:39:09.625 I was impressed. 1315 00:39:09.625 --> 00:39:11.250 Like, oh, does anything in the Rust 1316 00:39:11.250 --> 00:39:12.875 ecosystem support color emoji or is it 1317 00:39:12.875 --> 00:39:14.583 only a thing that browsers can do? 1318 00:39:14.583 --> 00:39:16.500 No, no, no, you can render emojis with 1319 00:39:16.500 --> 00:39:18.291 cosmic text, the Mac OS color ones. 1320 00:39:18.833 --> 00:39:20.500 Apparently the color emojis format for 1321 00:39:20.500 --> 00:39:22.166 Mac OS, Windows and Linux are completely 1322 00:39:22.166 --> 00:39:23.875 different because yeah, 1323 00:39:23.875 --> 00:39:25.291 that's why we can't have nice things. 1324 00:39:25.666 --> 00:39:27.666 But the demo of cosmic 1325 00:39:27.666 --> 00:39:29.833 text was actually pretty slow. 1326 00:39:30.166 --> 00:39:30.958 Like it's rendering a bunch 1327 00:39:30.958 --> 00:39:33.291 of text and you can select it. 1328 00:39:33.291 --> 00:39:34.250 It's kind of an editor. 1329 00:39:34.625 --> 00:39:38.000 But yeah, on this pretty good computer, I 1330 00:39:38.000 --> 00:39:39.833 expected it to be really fast, like running 1331 00:39:39.833 --> 00:39:42.083 at 60FPS or something and it did not. 1332 00:39:42.083 --> 00:39:43.125 So I don't know who's to blame here. 1333 00:39:43.291 --> 00:39:44.041 I don't know what's happening. 1334 00:39:44.583 --> 00:39:46.625 Cosmic text is probably great and fine 1335 00:39:46.625 --> 00:39:48.125 and this is not its fault. 1336 00:39:48.666 --> 00:39:49.791 But text rendering is just 1337 00:39:49.791 --> 00:39:51.333 more expensive than I thought. 1338 00:39:51.333 --> 00:39:54.375 I guess the trick is to not do it every 1339 00:39:54.375 --> 00:39:55.958 frame if you don't need to and like cache 1340 00:39:55.958 --> 00:39:56.666 as much as possible. 1341 00:39:57.083 --> 00:39:58.458 And I don't know, like, you know, 1342 00:39:58.458 --> 00:39:59.666 at-laces and whatnot. 1343 00:39:59.958 --> 00:40:00.916 And the demo is just 1344 00:40:00.916 --> 00:40:02.708 about correctness instead. 1345 00:40:02.958 --> 00:40:03.750 I have no idea. 1346 00:40:03.833 --> 00:40:05.500 But at this point, I started wondering, 1347 00:40:05.833 --> 00:40:08.000 well, I have demos on my website which 1348 00:40:08.000 --> 00:40:09.916 use more than just text. 1349 00:40:10.166 --> 00:40:11.041 Like what if I could just 1350 00:40:11.041 --> 00:40:12.750 render a whole browser in there? 1351 00:40:12.750 --> 00:40:14.041 Instead of messing around and having just 1352 00:40:14.041 --> 00:40:16.000 text and text renderers and Rust. 1353 00:40:16.458 --> 00:40:16.666 Okay. 1354 00:40:16.833 --> 00:40:18.041 When it comes to embedding browsers, 1355 00:40:18.250 --> 00:40:19.416 there are several options. 1356 00:40:19.666 --> 00:40:21.166 One of them is UltraLite, which is a 1357 00:40:21.166 --> 00:40:25.291 stripped down build of, I think, WebKit, 1358 00:40:25.833 --> 00:40:28.666 which means not Chrome, but like Safari 1359 00:40:29.000 --> 00:40:31.000 with like JavaScript core and whatnot. 1360 00:40:31.000 --> 00:40:32.791 It's this pure GPU HTML renderer. 1361 00:40:32.833 --> 00:40:33.875 I'm pretty sure that's, I 1362 00:40:33.875 --> 00:40:35.208 don't know how accurate that is. 1363 00:40:35.208 --> 00:40:36.291 It's GPU accelerated, 1364 00:40:36.291 --> 00:40:37.375 I'll acknowledge that. 1365 00:40:37.375 --> 00:40:39.208 There's SDKs for Rust, 1366 00:40:39.208 --> 00:40:41.500 non-official ones, but feasible. 1367 00:40:41.916 --> 00:40:43.000 But I took a look at the pricing. 1368 00:40:43.416 --> 00:40:44.291 James, what do you think this costs? 1369 00:40:44.625 --> 00:40:45.833 Wait, browsers cost money? 1370 00:40:45.916 --> 00:40:46.500 It's not a browser. 1371 00:40:46.791 --> 00:40:47.833 It's an embeddable 1372 00:40:47.833 --> 00:40:49.291 pure GPU HTML renderer. 1373 00:40:50.000 --> 00:40:50.500 HTML renderer? 1374 00:40:52.333 --> 00:40:53.375 Contact us for pricing? 1375 00:40:53.750 --> 00:40:55.250 No, it's a little better than that. 1376 00:40:55.250 --> 00:40:56.541 So there's a free tier up 1377 00:40:56.541 --> 00:40:59.250 until 100K of revenue per year. 1378 00:40:59.458 --> 00:41:00.750 And then there's a pro tier. 1379 00:41:00.916 --> 00:41:01.291 Okay. 1380 00:41:01.791 --> 00:41:02.958 You can pay something yearly. 1381 00:41:03.291 --> 00:41:03.750 Okay. 1382 00:41:03.791 --> 00:41:04.458 And it's 3K. 1383 00:41:05.291 --> 00:41:08.166 So some years flirt with 100K. 1384 00:41:08.541 --> 00:41:09.208 Wait, so where does the 1385 00:41:09.208 --> 00:41:10.250 browser come into this? 1386 00:41:11.125 --> 00:41:12.500 Where browser? 1387 00:41:12.833 --> 00:41:14.333 We were talking about video editing. 1388 00:41:14.708 --> 00:41:15.375 Where browser? 1389 00:41:15.791 --> 00:41:16.416 Where browser? 1390 00:41:16.708 --> 00:41:18.750 UltraLite lets you open a webpage pretty 1391 00:41:18.750 --> 00:41:21.541 much and render it in a GPU buffer. 1392 00:41:21.583 --> 00:41:23.500 You'd have a plugin that does browser 1393 00:41:23.500 --> 00:41:24.958 rendering, so you'd write JavaScript. 1394 00:41:25.500 --> 00:41:26.958 So you're running a scripting engine 1395 00:41:26.958 --> 00:41:29.125 inside of your plugin with a rendering 1396 00:41:29.125 --> 00:41:31.416 target, and then you're going splat into 1397 00:41:31.416 --> 00:41:32.916 the image and then like, you know, 1398 00:41:33.041 --> 00:41:34.000 rasterizing that into the 1399 00:41:34.000 --> 00:41:35.083 green screen or something? 1400 00:41:35.541 --> 00:41:37.041 Not sure I follow your explanation, but 1401 00:41:37.041 --> 00:41:38.083 yes, the idea would be 1402 00:41:38.083 --> 00:41:39.958 to-- It's probably wrong. 1403 00:41:40.708 --> 00:41:42.333 To have, instead of rendering text 1404 00:41:42.333 --> 00:41:45.416 directly, you would render web contents. 1405 00:41:45.750 --> 00:41:47.125 And that way, those motion 1406 00:41:47.125 --> 00:41:48.125 design tools are available. 1407 00:41:48.333 --> 00:41:51.041 I forgot their names, but you can write 1408 00:41:51.041 --> 00:41:53.083 TypeScript code essentially instead of 1409 00:41:53.083 --> 00:41:54.958 doing things on a timeline and say, okay, 1410 00:41:54.958 --> 00:41:56.333 this element comes in, this element 1411 00:41:56.333 --> 00:41:57.083 rotates, this 1412 00:41:57.083 --> 00:41:59.583 element's opacity does this. 1413 00:41:59.625 --> 00:42:01.750 Yeah, code your own visualizations. 1414 00:42:02.458 --> 00:42:04.250 And then the idea would be, instead of 1415 00:42:04.250 --> 00:42:06.000 exporting that, like rendering all that 1416 00:42:06.000 --> 00:42:08.375 to video, you could have an OpenFX plugin 1417 00:42:08.375 --> 00:42:10.333 where you're just, hey, here's the file 1418 00:42:10.333 --> 00:42:12.333 for that scene or whatever. 1419 00:42:12.333 --> 00:42:13.458 It would render frames 1420 00:42:13.458 --> 00:42:15.875 as requested by the host. 1421 00:42:16.208 --> 00:42:16.458 Interesting. 1422 00:42:16.958 --> 00:42:17.958 Using an HTML engine. 1423 00:42:18.416 --> 00:42:18.625 Yeah. 1424 00:42:18.958 --> 00:42:20.250 But HTML engine is, you know. 1425 00:42:20.250 --> 00:42:22.041 That's how the question has become, how 1426 00:42:22.041 --> 00:42:24.125 do we embed HTML content? 1427 00:42:24.333 --> 00:42:26.000 So how do we render that onto a canvas? 1428 00:42:27.000 --> 00:42:28.666 So UltraLite, easy to embed in 1429 00:42:28.666 --> 00:42:31.458 everything, but I'm not paying 3K a year, 1430 00:42:31.750 --> 00:42:33.791 the year that I break 100K of revenue. 1431 00:42:34.125 --> 00:42:36.416 I looked into, it was a nice, like. 1432 00:42:36.458 --> 00:42:37.000 That was my question. 1433 00:42:37.458 --> 00:42:38.333 Servo was my next question. 1434 00:42:38.666 --> 00:42:39.458 Yeah, Servo. 1435 00:42:39.500 --> 00:42:41.083 Servo has a gazillion dependencies. 1436 00:42:41.583 --> 00:42:43.625 It has SpiderMonkey, it has 1437 00:42:43.625 --> 00:42:45.625 its own build system called "mach". 1438 00:42:46.416 --> 00:42:48.250 I tried to build it and then immediately 1439 00:42:48.250 --> 00:42:49.291 complain about GStreamer. 1440 00:42:49.750 --> 00:42:52.708 I freaking hate slash love GStreamer. 1441 00:42:53.166 --> 00:42:54.875 GStreamer is one of those words that I've 1442 00:42:54.875 --> 00:42:56.875 seen it install logs and stuff like that. 1443 00:42:56.875 --> 00:42:58.666 I have no idea how it works and it would 1444 00:42:58.666 --> 00:42:59.833 probably hurt my head to learn. 1445 00:43:01.125 --> 00:43:02.125 (Laughing) It's not that bad. 1446 00:43:02.125 --> 00:43:03.166 I should do an episode about it. 1447 00:43:03.166 --> 00:43:05.250 But it has a gazillion dependencies and 1448 00:43:05.250 --> 00:43:06.583 libraries spread all over the place 1449 00:43:06.583 --> 00:43:07.958 because they lean into 1450 00:43:07.958 --> 00:43:09.750 dynamic linking a lot. 1451 00:43:10.041 --> 00:43:13.375 The Servo embedding story is before it 1452 00:43:13.375 --> 00:43:16.000 was abandoned, someone was keeping a 1453 00:43:16.000 --> 00:43:17.208 repository on GitHub with 1454 00:43:17.208 --> 00:43:19.208 examples, Paul Rouget, I think. 1455 00:43:19.500 --> 00:43:21.166 And that hasn't been 1456 00:43:21.166 --> 00:43:22.000 touched in six years. 1457 00:43:22.500 --> 00:43:24.791 In the meantime, it's been unabandoned 1458 00:43:25.000 --> 00:43:26.625 and they have collaborated 1459 00:43:26.625 --> 00:43:29.583 with Tauri in January of 2024. 1460 00:43:30.041 --> 00:43:31.458 Tauri, they're like Electron but in Rust. 1461 00:43:32.291 --> 00:43:33.416 And they've done a proof 1462 00:43:33.416 --> 00:43:34.291 of concept integration. 1463 00:43:34.708 --> 00:43:35.583 It got picked up by that 1464 00:43:35.583 --> 00:43:36.791 collective, what are they called? 1465 00:43:37.125 --> 00:43:37.458 Igalia. 1466 00:43:37.958 --> 00:43:38.583 Igalia, yeah, yeah, yeah. 1467 00:43:38.583 --> 00:43:40.125 It got picked up by Igalia because a lot 1468 00:43:40.125 --> 00:43:41.625 of people want a little browser engine. 1469 00:43:41.625 --> 00:43:42.375 I guess what you were 1470 00:43:42.375 --> 00:43:43.291 saying makes a lot of sense. 1471 00:43:43.458 --> 00:43:45.750 I've seen people wanting to embed it in 1472 00:43:45.750 --> 00:43:47.958 either VR, XR type stuff, 1473 00:43:47.958 --> 00:43:49.166 or even in game engines. 1474 00:43:49.541 --> 00:43:50.625 Because like you said, you want someone 1475 00:43:50.625 --> 00:43:52.791 to be able to write, use web tech to make 1476 00:43:52.791 --> 00:43:56.291 your heads up display in a VR setting or 1477 00:43:56.291 --> 00:43:56.875 something like that. 1478 00:43:56.875 --> 00:43:58.291 So actually it does kind of track. 1479 00:43:58.791 --> 00:44:01.833 A lot of people do, but they use CEF, 1480 00:44:02.083 --> 00:44:02.958 which I'm just getting to. 1481 00:44:03.250 --> 00:44:04.500 I just wanted to mention that like the 1482 00:44:04.500 --> 00:44:07.125 Tauri embedding proof of concept thingy 1483 00:44:07.125 --> 00:44:08.875 happened in January of 2024. 1484 00:44:09.250 --> 00:44:11.291 And then in February of 2025, there's a 1485 00:44:11.291 --> 00:44:12.625 new blog article with a tagline 1486 00:44:12.625 --> 00:44:13.500 of "Servo is becoming 1487 00:44:13.500 --> 00:44:15.291 truly embeddable this year." 1488 00:44:15.291 --> 00:44:16.541 So, you know, we're 1489 00:44:16.541 --> 00:44:17.625 halfway through the year. 1490 00:44:17.625 --> 00:44:19.000 I'm excited to see what's to come. 1491 00:44:19.000 --> 00:44:20.708 But for now in the documentation, there's 1492 00:44:20.708 --> 00:44:22.583 nothing about how to embed or anything. 1493 00:44:23.333 --> 00:44:24.750 The year of servo on the embedded. 1494 00:44:25.916 --> 00:44:28.416 Speaking of embedded, what most games and 1495 00:44:28.416 --> 00:44:30.458 like, for example, the League of Legends 1496 00:44:30.458 --> 00:44:32.666 clients, Riot Games client and all that, 1497 00:44:32.916 --> 00:44:34.666 all the applications that are mostly 1498 00:44:34.666 --> 00:44:37.708 Electron, but actually not, they're using 1499 00:44:37.708 --> 00:44:40.375 the Chromium embedded framework, CEF. 1500 00:44:40.375 --> 00:44:43.208 There's a crate for that, which entire 1501 00:44:43.208 --> 00:44:44.875 page just says, use the 1502 00:44:44.875 --> 00:44:46.291 Chromium embedded framework in Rust. 1503 00:44:46.583 --> 00:44:47.208 And that's it. 1504 00:44:47.208 --> 00:44:47.958 I love it. 1505 00:44:47.958 --> 00:44:48.791 Straight into the point. 1506 00:44:48.833 --> 00:44:49.000 Yes. 1507 00:44:49.250 --> 00:44:50.250 With a version number 1508 00:44:50.250 --> 00:44:51.625 that definitely makes sense. 1509 00:44:52.375 --> 00:44:54.208 It matches the Chromium version. 1510 00:44:54.458 --> 00:44:55.958 You can tell it's automated because it 1511 00:44:55.958 --> 00:44:57.166 was less published three days ago. 1512 00:44:57.375 --> 00:44:58.875 So I'm like, and it works, right? 1513 00:44:58.875 --> 00:45:00.375 I tried their example and it works. 1514 00:45:00.375 --> 00:45:01.333 You can have your own 1515 00:45:01.333 --> 00:45:03.291 little browser with no Chrome. 1516 00:45:03.583 --> 00:45:05.791 Chrome is the browser interface, like the 1517 00:45:05.791 --> 00:45:08.708 address bar, next previous button, the 1518 00:45:08.708 --> 00:45:09.583 menus and everything. 1519 00:45:09.583 --> 00:45:10.333 That's the Chrome. 1520 00:45:10.750 --> 00:45:11.500 Firefox has a Chrome. 1521 00:45:11.916 --> 00:45:13.250 It's written in CSS and JavaScript. 1522 00:45:13.458 --> 00:45:16.125 But I don't know much about CEF and that 1523 00:45:16.125 --> 00:45:18.875 simple demo worked, but there's an 1524 00:45:18.875 --> 00:45:19.875 offscreen rendering demo 1525 00:45:19.875 --> 00:45:20.875 and that one didn't work. 1526 00:45:21.166 --> 00:45:23.875 So I'm assuming it was coded against 1527 00:45:23.875 --> 00:45:25.041 Windows or something. 1528 00:45:25.041 --> 00:45:27.791 And you need this whole kind of app 1529 00:45:27.791 --> 00:45:29.750 bundle structure with all the libraries 1530 00:45:29.750 --> 00:45:30.541 in the right places. 1531 00:45:30.958 --> 00:45:32.708 It seems like it's a lot of effort. 1532 00:45:33.208 --> 00:45:35.291 I just, I gave up on that avenue. 1533 00:45:35.625 --> 00:45:37.541 I'm sure it's probably the right way to 1534 00:45:37.541 --> 00:45:39.875 go at this point because when I look at 1535 00:45:39.875 --> 00:45:41.833 Servo, for example, I load it, my voice 1536 00:45:41.833 --> 00:45:42.708 out in it to see it. 1537 00:45:42.708 --> 00:45:43.916 And it supports a lot of things. 1538 00:45:44.125 --> 00:45:47.291 But for example, I have SVG diagrams that 1539 00:45:47.291 --> 00:45:50.500 use HTML text formatting and it only 1540 00:45:50.500 --> 00:45:52.250 showed the shapes and it didn't show the 1541 00:45:52.250 --> 00:45:54.000 text because it didn't support like 1542 00:45:54.000 --> 00:45:56.166 foreign objects, nodes in 1543 00:45:56.166 --> 00:45:57.625 SVG, which is kind of a bummer. 1544 00:45:58.541 --> 00:46:00.875 Finally, I figured the easiest way to 1545 00:46:00.875 --> 00:46:02.583 experiment with all this would be to use 1546 00:46:02.583 --> 00:46:05.375 Electron because Electron has an 1547 00:46:05.375 --> 00:46:06.875 offscreen rendering API and you can just 1548 00:46:06.875 --> 00:46:09.375 write a bit of JavaScript code. 1549 00:46:09.875 --> 00:46:13.083 And the way I started experimenting with 1550 00:46:13.083 --> 00:46:16.375 this is just listening on some TCP port 1551 00:46:16.375 --> 00:46:18.291 for some RPC requests, 1552 00:46:18.875 --> 00:46:20.375 which is just a line of JSON. 1553 00:46:20.708 --> 00:46:22.833 There's a bunch of ways to capture the 1554 00:46:22.833 --> 00:46:24.375 contents of the webpage and one of them 1555 00:46:24.375 --> 00:46:25.625 is CPU, the other is GPU. 1556 00:46:26.083 --> 00:46:28.875 I only tried the CPU because GPU is 1557 00:46:28.875 --> 00:46:30.250 essentially you get a shared texture 1558 00:46:30.250 --> 00:46:32.041 handle and then you need to open it and 1559 00:46:32.041 --> 00:46:34.875 like copy from the, or try not to copy 1560 00:46:34.875 --> 00:46:36.416 from the OpenFX plugin. 1561 00:46:36.916 --> 00:46:39.291 I'm pretty sure there would be a way to, 1562 00:46:40.000 --> 00:46:42.000 instead of filling out the buffer, say, 1563 00:46:42.000 --> 00:46:45.000 hey, here's the shared texture handle, go 1564 00:46:45.000 --> 00:46:47.125 take that, go use that, I don't know. 1565 00:46:47.541 --> 00:46:49.375 Yeah, it gives those nice images. 1566 00:46:50.125 --> 00:46:52.375 It's, this is not different from all the 1567 00:46:52.375 --> 00:46:53.875 other slides, but that one was generated 1568 00:46:53.875 --> 00:46:55.291 with the whole Electron pipeline. 1569 00:46:55.500 --> 00:46:56.958 And I figured, yeah, okay, the plugin 1570 00:46:56.958 --> 00:46:58.625 could like start Electron in the 1571 00:46:58.625 --> 00:46:59.958 background, send those RPC 1572 00:46:59.958 --> 00:47:02.375 requests, get the images back. 1573 00:47:02.375 --> 00:47:03.875 But do you have any idea how long it 1574 00:47:03.875 --> 00:47:05.208 takes to generate one 1575 00:47:05.208 --> 00:47:06.833 frame, one 4K frame? 1576 00:47:07.375 --> 00:47:08.166 I bet it's not gonna be 1577 00:47:08.166 --> 00:47:09.083 three to four milliseconds. 1578 00:47:09.791 --> 00:47:10.875 It's not three to four milliseconds. 1579 00:47:11.833 --> 00:47:12.125 Okay. 1580 00:47:13.750 --> 00:47:14.416 650 milliseconds. 1581 00:47:15.333 --> 00:47:17.541 Okay, love that, love that. 1582 00:47:17.583 --> 00:47:20.166 So in that there's like Node.js startup 1583 00:47:20.166 --> 00:47:21.791 time connecting over TCP 1584 00:47:21.791 --> 00:47:23.250 socket, this is not a good test. 1585 00:47:23.250 --> 00:47:25.666 And also it's encoding everything to PNG. 1586 00:47:26.458 --> 00:47:27.958 So you can definitely shave 1587 00:47:27.958 --> 00:47:30.000 off a bunch of times here. 1588 00:47:30.000 --> 00:47:30.750 There's next steps, 1589 00:47:31.250 --> 00:47:32.250 more research is required. 1590 00:47:32.958 --> 00:47:34.291 And next season is 1591 00:47:34.291 --> 00:47:36.208 required for this for sure. 1592 00:47:36.416 --> 00:47:38.666 Because I think the shared texture handle 1593 00:47:38.666 --> 00:47:41.500 thing is the way to go if doing Electron. 1594 00:47:41.833 --> 00:47:43.958 I think embedding servo is probably a lot 1595 00:47:43.958 --> 00:47:46.166 better, but then I started thinking 1596 00:47:46.166 --> 00:47:48.625 bigger picture, even if everything worked 1597 00:47:48.625 --> 00:47:50.791 well and was low latency, like you get 1598 00:47:50.791 --> 00:47:52.125 real time playback and everything. 1599 00:47:52.125 --> 00:47:54.750 You now have some interface inside the 1600 00:47:54.750 --> 00:47:56.000 host, inside DaVinci Resolve. 1601 00:47:56.500 --> 00:47:58.625 And ideally you'd also want to be able 1602 00:47:58.625 --> 00:48:00.416 to, if you're doing motion design, if you 1603 00:48:00.416 --> 00:48:02.208 have like bits of code, like you can zoom 1604 00:48:02.208 --> 00:48:04.166 in and highlight and whatnot, that's 1605 00:48:04.166 --> 00:48:05.083 something you need to do 1606 00:48:05.083 --> 00:48:07.458 inside of your web contents. 1607 00:48:07.916 --> 00:48:09.875 Right, you can't do it in the venture 1608 00:48:09.875 --> 00:48:12.041 resolve because if you zoom in, if you 1609 00:48:12.041 --> 00:48:13.416 have a clip, you put it in your scene and 1610 00:48:13.416 --> 00:48:14.166 then you zoom that in, 1611 00:48:14.250 --> 00:48:15.666 it's gonna be pixelated. 1612 00:48:16.291 --> 00:48:16.958 All right, does that make sense? 1613 00:48:17.541 --> 00:48:18.250 Yeah, yeah, yeah. 1614 00:48:18.291 --> 00:48:20.416 You're doing like pan and scan on a 1615 00:48:20.416 --> 00:48:22.000 rasterized image, which means either the 1616 00:48:22.000 --> 00:48:24.583 rasterized image needs to be huge, so you 1617 00:48:24.583 --> 00:48:26.291 can zoom into whatever fidelity you want, 1618 00:48:26.708 --> 00:48:29.000 or you want it rendered on the fly, which 1619 00:48:29.000 --> 00:48:31.125 is more like a camera flying around. 1620 00:48:31.916 --> 00:48:32.125 Exactly. 1621 00:48:32.458 --> 00:48:35.000 So that means now that all your 1622 00:48:35.000 --> 00:48:36.875 positioning, your zooming, everything 1623 00:48:36.875 --> 00:48:39.708 needs to be a part, that needs to be a 1624 00:48:39.708 --> 00:48:42.541 parameter, that you can keyframe inside 1625 00:48:42.541 --> 00:48:44.541 the host, or it needs to be part of a 1626 00:48:44.541 --> 00:48:46.875 scene that you can kind of open in a 1627 00:48:46.875 --> 00:48:49.000 separate editor, a web-based editor, and 1628 00:48:49.000 --> 00:48:51.833 then the OpenFX plugin is just a 1629 00:48:51.833 --> 00:48:54.041 renderer, and it's just like giving you a 1630 00:48:54.041 --> 00:48:55.416 frame of a predetermined scene. 1631 00:48:55.625 --> 00:48:56.875 And you still have the problem that I 1632 00:48:56.875 --> 00:48:59.458 often have, which is you usually want to 1633 00:48:59.458 --> 00:49:01.083 do the animation synchronized with some 1634 00:49:01.083 --> 00:49:03.208 audio, right, someone's talking, and then 1635 00:49:03.208 --> 00:49:05.250 you want the code to come in as they talk 1636 00:49:05.250 --> 00:49:06.250 about different things. 1637 00:49:06.250 --> 00:49:09.250 And if your editor is outside of resolve, 1638 00:49:09.541 --> 00:49:10.250 you're not gonna get the 1639 00:49:10.250 --> 00:49:11.750 audio from the resolve timeline. 1640 00:49:12.166 --> 00:49:14.041 At that point, the only reasonable course 1641 00:49:14.041 --> 00:49:15.416 of action is to make your 1642 00:49:15.416 --> 00:49:17.666 own nonlinear editor, right? 1643 00:49:18.583 --> 00:49:19.291 I was wondering if it 1644 00:49:19.291 --> 00:49:20.291 was gonna come to this. 1645 00:49:20.375 --> 00:49:21.375 Well, I certainly 1646 00:49:21.375 --> 00:49:22.666 haven't done it in a day. 1647 00:49:23.291 --> 00:49:25.458 I'm not planning on doing it because how 1648 00:49:25.458 --> 00:49:27.500 do you hire editors for your own 1649 00:49:27.500 --> 00:49:29.583 nonlinear editor software? 1650 00:49:30.333 --> 00:49:32.500 But that would be the way to do it. 1651 00:49:33.375 --> 00:49:34.958 You sound like you're standing on the 1652 00:49:34.958 --> 00:49:36.875 ledge, and I'm not sure whether to slowly 1653 00:49:36.875 --> 00:49:38.958 talk you down or not, or if you're 1654 00:49:38.958 --> 00:49:41.750 excited for this jump into the unknown. 1655 00:49:42.333 --> 00:49:44.083 On top of Gear Acquisition Syndrome, I 1656 00:49:44.083 --> 00:49:46.375 have Software Replacement Syndrome. 1657 00:49:46.583 --> 00:49:48.166 I use software too much. 1658 00:49:49.333 --> 00:49:50.291 DaVinci Resolve used to be this 1659 00:49:50.291 --> 00:49:52.333 really intimidating, big piece of 1660 00:49:52.333 --> 00:49:53.375 software that I didn't 1661 00:49:53.375 --> 00:49:54.875 know how to deal with. 1662 00:49:55.166 --> 00:49:58.666 And now I kind of see the seams of the 1663 00:49:58.666 --> 00:49:59.833 different parts of the 1664 00:49:59.833 --> 00:50:01.583 things that they pulled together. 1665 00:50:02.041 --> 00:50:06.041 And I think you could do better if you 1666 00:50:06.041 --> 00:50:06.875 started from scratch. 1667 00:50:07.000 --> 00:50:08.583 The problem is, of course, all the 1668 00:50:08.583 --> 00:50:10.583 unknown unknowns, all the things that 1669 00:50:10.583 --> 00:50:12.208 they got absolutely right, and I have no 1670 00:50:12.208 --> 00:50:14.291 idea how to do high-performance image 1671 00:50:14.291 --> 00:50:15.500 processing using shaders. 1672 00:50:16.333 --> 00:50:16.958 No idea about that. 1673 00:50:17.666 --> 00:50:19.583 Or just 30 years of paper cut shaving. 1674 00:50:20.291 --> 00:50:21.875 There's a huge amount of, with Second 1675 00:50:21.875 --> 00:50:23.208 System Syndrome, where you're like, "I 1676 00:50:23.208 --> 00:50:24.333 could make it so much easier." 1677 00:50:24.708 --> 00:50:27.250 And you realize that so much of the 1678 00:50:27.250 --> 00:50:29.375 things that make it painful are like, 1679 00:50:29.375 --> 00:50:31.500 "Oh, we decided it was worth that pain, 1680 00:50:31.875 --> 00:50:33.583 accumulated over 30 years." 1681 00:50:33.583 --> 00:50:34.250 You know what I mean? 1682 00:50:34.958 --> 00:50:36.041 But yeah. 1683 00:50:36.291 --> 00:50:37.833 You cheat by doing less, so if you could 1684 00:50:37.833 --> 00:50:39.000 figure out how to do something that only 1685 00:50:39.000 --> 00:50:40.916 matters for you, hell yeah. 1686 00:50:41.416 --> 00:50:43.333 But yeah, I imagine 1687 00:50:43.333 --> 00:50:44.250 it's gonna be a thing. 1688 00:50:45.833 --> 00:50:47.875 Well, I'm gonna try not to 1689 00:50:47.875 --> 00:50:49.625 go any further down that road. 1690 00:50:50.208 --> 00:50:50.666 Wish me luck. 1691 00:50:50.875 --> 00:50:51.583 Famous last words. 1692 00:50:52.000 --> 00:50:52.958 I think we both know. 1693 00:50:54.125 --> 00:50:55.250 That road does not end. 1694 00:50:55.708 --> 00:50:57.041 I was gonna say, we are both the epitome 1695 00:50:57.041 --> 00:50:58.541 of if you give a mouse a cookie. 1696 00:50:59.083 --> 00:51:01.250 We figured out scripting to, okay, well, 1697 00:51:01.250 --> 00:51:01.833 I guess I need to 1698 00:51:01.833 --> 00:51:02.875 replace DaVinci Resolve. 1699 00:51:03.458 --> 00:51:04.833 (Laughs) 1700 00:51:04.916 --> 00:51:07.875 But I, no, no, I-- I will be strong. 1701 00:51:08.291 --> 00:51:09.333 All right, I believe in you. 1702 00:51:09.708 --> 00:51:10.333 Thank you. 1703 00:51:10.333 --> 00:51:11.541 I'm looking forward to next season. 1704 00:51:12.625 --> 00:51:12.875 (Laughs) 1705 00:51:13.166 --> 00:51:13.500 Me too. 1706 00:51:21.458 --> 00:51:22.958 This episode is sponsored by Depot, 1707 00:51:23.125 --> 00:51:23.916 the build acceleration 1708 00:51:23.916 --> 00:51:25.208 platform that's on a mission 1709 00:51:25.208 --> 00:51:26.833 to make all builds near instant. 1710 00:51:27.208 --> 00:51:28.541 If you're tired of watching your builds 1711 00:51:28.541 --> 00:51:29.750 and GitHub actions crawl 1712 00:51:29.750 --> 00:51:31.041 like the modern day equivalent 1713 00:51:31.041 --> 00:51:33.041 of paint drying, give Depot's GitHub 1714 00:51:33.041 --> 00:51:34.083 Actions runners a try. 1715 00:51:34.541 --> 00:51:35.958 They're up to 10 times faster with 1716 00:51:35.958 --> 00:51:37.208 unlimited concurrency, 1717 00:51:37.250 --> 00:51:38.583 faster caching, support for 1718 00:51:38.583 --> 00:51:40.083 Linux, macOS and Windows, 1719 00:51:40.375 --> 00:51:41.291 and they plug right into 1720 00:51:41.291 --> 00:51:42.375 other Depot optimizations 1721 00:51:42.625 --> 00:51:44.333 like accelerated container image builds 1722 00:51:44.458 --> 00:51:45.416 and remote caching for 1723 00:51:45.416 --> 00:51:47.625 Bazel, Turborepo, Gradle and more. 1724 00:51:48.041 --> 00:51:49.541 Depot is built by developers who are 1725 00:51:49.541 --> 00:51:50.541 tired of wasting time 1726 00:51:50.541 --> 00:51:52.041 waiting on builds instead of shipping. 1727 00:51:52.333 --> 00:51:53.125 It's made for teams 1728 00:51:53.125 --> 00:51:54.125 that wanna move faster 1729 00:51:54.375 --> 00:51:55.083 and stay focused on 1730 00:51:55.083 --> 00:51:55.958 what actually matters. 1731 00:51:56.458 --> 00:51:57.125 That's why companies 1732 00:51:57.125 --> 00:51:58.583 like PostHog use Depot 1733 00:51:58.583 --> 00:52:00.458 to cut build times from over three hours 1734 00:52:00.458 --> 00:52:01.583 to just three minutes, 1735 00:52:01.833 --> 00:52:02.916 saving tens of thousands 1736 00:52:02.916 --> 00:52:04.250 of build hours every week. 1737 00:52:04.541 --> 00:52:05.500 Start your free seven 1738 00:52:05.500 --> 00:52:06.875 day trial at depot.dev 1739 00:52:06.875 --> 00:52:08.083 and let them know we sent you.