4/21/2025 - Exploring the Role of AI in Creative Storytelling and Film Production

Description

Chapters

Transcript

0:00 Little doggy, are you coming with
0:04 [Music]
0:20 daddy? Sh. Quiet. Quiet for a second.
0:24 [Music]
0:27 [Applause]
0:29 [Music]
0:35 [Music]
0:55 Oh, freedom came my way. That
0:57 [Music]
1:03 night just like a jet
1:06 plane in and out of sight.
1:10 I was hauling ass at a million miles an
1:15 hour wondering how
1:18 [Music]
1:21 hard when they came into the
1:28 station. They said I was bad beyond
1:33 repair, but I got no qualms with my
1:38 situation.
1:41 Say here I am.
1:47 [Music]
1:50 Say
1:52 sh won't you dare to say
1:58 sh won't you dare to say
2:06 sh. Uhhuh.
2:09 Yeah. Leave a message. Add your number.
2:13 Please.
2:16 Take a time to want to satisfy me.
2:20 [Music]
2:22 Take all these old
2:24 fantasies. Send them care of me.
2:29 [Music]
2:33 Happy Monday, good people. What is
2:35 happening out there in Tik Tok land? Tik
2:37 Tok's still
2:38 here. I don't know what loser suggested
2:41 we wouldn't be here
2:43 today, but we're here. We're
2:49 here. Got to stop calling myself a
2:51 loser.
2:58 [Music]
3:26 Why you getting so Rammy? Just because I
3:27 was Ramy? Because I'm Ramy? You're Ramy?
3:30 Someone's got to break the cycle,
3:32 champ. Someone's got to break the
3:35 cycle. Someone comes at you, they're
3:37 Ramy. You have two choices. You get
3:40 Ramy or you choose to back it
3:44 down. Let's take a breath. Let's calm it
3:48 down. And champ, what did you choose?
3:51 Not the higher road. Oh, yeah. Run to
3:54 your bed. Run to your bed. It's what
3:56 happens. It's what happens. All of our
3:58 fights end like
4:00 this. He just jumps up on his bed and
4:02 lays down, pretends like he's a dog.
4:04 It's like, "Yeah, whatever. I'm going to
4:06 let this pass as if I don't have
4:11 language." Show off. You know, there's
4:14 got to be something said for not having
4:17 language cuz these brains [ __ ] us up.
4:20 You know, artificial intelligence,
4:22 everyone's like, "Oh, they hallucinate.
4:24 Have you seen a human?
4:27 [Music]
4:31 I know. Once they can talk, once dogs
4:34 can talk, they'll be like, "H, just
4:37 stop. Just give me chicken. Just give me
4:40 Could I have some cheese? Could I Yeah.
4:42 Yes, I'll sit. I'll I'll do whatever the
4:44 [ __ ] you want. Could I please just have
4:46 some
4:49 cheese?"
4:51 Right. They're going to be like, "Are
4:52 you're going to put a bow? You're going
4:54 to put a bow on me and then we're going
4:55 to walk
4:56 outside. That's That's how we're playing
4:59 this. That's what I have to do for
5:01 cheese. It's a miserable existence. You
5:03 know, I used to be a wolf, right?
5:09 [Music]
5:51 H I wonder what the first word ever
5:54 spoken was. That's actually a really
5:56 interesting
5:57 thought. I often think about that
5:59 because you I mean you know Steo it
6:02 wasn't like
6:04 ugg hey hey Bob like it didn't it was
6:08 just like ugg and then someone went
6:13 u like it was it was a slightly modified
6:16 ugg right and then someone else
6:20 went bug bug maybe it was bug you're
6:25 like ugg bug
6:27 [Laughter]
6:36 You know, it wasn't as good as you'd
6:38 make it out to be. You'd be like, I'd
6:39 give anything to be at that moment. And
6:41 then you AI [ __ ] solves time travel
6:43 and you're like, I want to go back to
6:44 when the first word happened. AI's like,
6:47 all right, you're going to be
6:48 disappointed. You're like, no, I'm not.
6:49 You should be like, uh,
6:53 uh, huh?
6:59 And then when whoever that was spoke
7:01 that word, all the other ones were
7:04 like, "He he's he's off his [ __ ]
7:07 rocker." Like the the ug version of
7:09 that. They all just went
7:11 ah and they just walked
7:14 away. He was sitting there like bug bug
7:17 bug bug. And then like a a mosquito
7:21 lands on his nose, stings. He was like,
7:23 "Bug."
7:24 [Laughter]
7:30 Huh? Yeah. Imagining moments in history
7:33 here at the AI learning lab. We don't
7:35 need no stinking AI. We'll just make it
7:38 up.
7:41 [Laughter]
7:43 [Music]
7:53 There's been something, baby, I've been
7:55 trying to
7:57 say for an age and it seems I don't know
8:02 how. Past in a future now surrounding
8:05 me. You know, I bet I bet some of the
8:08 first words are like ma and pa because
8:10 like the roots for mother and father are
8:13 similar across many many languages.
8:15 Yeah. Oh, that's
8:16 look Brandon just said this studies of
8:19 ancient language roots suggest the first
8:21 meaningful words were ma or mama. Yeah,
8:23 that makes sense, right? You keep seeing
8:25 this person, you're like, oh, you know
8:28 what it was? They're like they get a
8:30 little bit of milk from the from the,
8:31 you know, from the source and they're
8:34 like m
8:37 M
8:39 mama. They're like They're like more.
8:42 Mama stands for more in
8:45 [Laughter]
8:47 [Music]
8:52 Caveman. Oh, this is so out of tune.
8:54 It's not
8:55 [Music]
8:59 playable. All right, whatever. I won't
9:01 tort you with that anymore because
9:03 that's ungodly
9:06 sounding. All right. Fantastic.
9:09 Fantastic. Dear Bob. Hey Bob. Wonder if
9:12 Bob was the first
9:15 word. Bob. Bob.
9:19 Bob. Bob. Tell him what he's won. It
9:23 probably just came out like
9:28 that. Oh, good lord. Good lord. Good
9:32 lordy. Good lordy. Good lordy. All right
9:37 fantastic man. Turn
9:39 into we haven't invented prizes yet.
9:42 There's some caveman walking around. Bob
9:44 tell him what he's won. People are like
9:46 Bob. Who's Bob? What is
9:54 one? Didn't we make a caveman Kyle once?
9:57 We might have.
9:59 Oh, this reminds me of my great aunt's
10:02 favorite joke. Oh,
10:03 [Laughter]
10:07 boy. So, I think I have an idea for
10:10 something that could be fun tonight. So,
10:12 we'll we'll play with that. I might get
10:14 if if anyone wants me to, I'm happy to
10:16 go on a Monday
10:18 meltdown, you know, ask me something
10:20 innocent about AI and I'll [ __ ] take
10:23 off. I'll [ __ ] take off on it.
10:27 I had an interesting talk today with
10:29 um couple of my my favorite employees uh
10:33 from agency.com back in the olden timey
10:36 days of the web. Four of us got
10:38 together. It was Ben Jones and Laurent
10:40 Stanovich and John Knack. John Knack
10:42 works for Microsoft now. And Ben's got
10:45 his own agency. He worked at Google for
10:47 a bunch of years. And uh and Laurent's
10:50 working at a uh a digital marketing
10:52 agency. and they're all just [ __ ]
10:55 brilliant. And
10:57 uh I was reminded that I've still got
11:00 that annoying
11:05 optimism. It was funny. They were they
11:08 were a little kermoginly about AI.
11:10 John's a little bitter about, you know,
11:12 life at Microsoft. I think Ben's, you
11:15 know, just in the middle of a startup
11:17 [Laughter]
11:19 grind. Oh, man. Like the new glasses.
11:23 Thank you very much, sir. Appreciate
11:25 that. These are uh these are
11:28 Zeni Zeni mail order
11:30 [Laughter]
11:33 glasses. You want to know something
11:36 stupid? I shouldn't admit. I found it.
11:39 You found what? On the screen. Where is
11:43 it? I I'm sharing my screen. Oh,
11:48 okay. Fire Learning Lab. That's good.
11:53 Yeah. Yeah. Actually, that's good. Back
11:55 then, I'd have been like, "Oh, this fire
11:57 thing is going to be big." People are
11:58 like, "No, just go go get water. Just
12:01 get another
12:02 blanket. Fire dangerous. Fire bad. Fire
12:06 is going to kill us all." Like, "No, no,
12:08 fire's going to be good." Like, imagine
12:09 if we tame fire and put it in like a pit
12:11 and then we invent these things called
12:13 marshmallows. You stick them on the
12:14 fire. No, no, fire bad.
12:19 How would I get started in short film
12:21 production using chat GPT? That's that's
12:24 actually a really great question. In
12:25 fact, so what I was going to do tonight,
12:28 hang on a
12:29 sec. Let me show you. I'm going to go to
12:32 go to my little uh it's called
12:36 mymind.com and it's a little like I
12:38 don't know what are they? It's like
12:39 Pinterest but for yourself. if you just
12:41 stick [ __ ] in there. And I use it mostly
12:43 to to gather
12:46 um uh srefs for midjourney, all these
12:50 different styles, right? And but this
12:55 one was a post that I saw right before I
12:59 went live tonight and it's about
13:01 creating character sheets. So So Chat
13:04 GPT is now really good at consistent
13:08 characters.
13:10 Um it does not have the sophistication
13:13 of midjourney
13:15 but I would give up well I am finding
13:19 myself giving up sophistication for
13:23 consistency and chatbt also does it can
13:26 mimic styles it can do typography really
13:28 well so if you want a billboard that
13:30 says you know tell them what they've won
13:31 Bob you can have a billboard that says
13:33 that you don't have to spend four hours
13:35 [ __ ] you know trying to prompt
13:37 something into existence or photoshop
13:39 stopping it.
13:41 Um, but let me let me give you the the
13:45 the basic rundown
13:48 on how to get started on short film
13:51 production with Chat
13:54 GPT.
13:56 Um, everything starts with story. Like I
13:59 don't I don't know where you are in
14:01 the in the experience thing. If you're
14:04 like, "Yeah, I've been making movies for
14:06 40 years. cuz I just needed to know how
14:07 to do it in AI [ __ ] Like if you're
14:09 that then this some of this stuff is
14:12 going to be ridiculous.
14:13 But you've got to start with story.
14:16 You've got to start with a story,
14:19 right? What's cool about AI is that you
14:22 can tell big stories, you can tell small
14:24 stories.
14:26 Um, and so where I would start is start
14:30 with a very small story. Start with a
14:32 story of like a single character, a
14:34 single location, and like a 30-se
14:37 secondond narrative arc, right? Like a I
14:40 don't know, a squirrel drops a peanut,
14:42 discovers a bunch of other peanuts, and
14:44 then says, "Who buried these?" Right?
14:46 Something something just silly, stupid.
14:51 But you can give chatbt an idea of of
14:55 what you want and then say, "Give me 20
14:57 ideas that fit that format." So, a one
15:01 minute film, two characters or or
15:03 whatever it might be or or something
15:04 that puts out a message that you want to
15:07 tell. So, start with
15:10 ideiation and then if you're going to
15:12 tell something of decent length, like a
15:14 minute or more, then you're going to
15:16 want to outline it. And even if you're
15:18 going to do something really short, it's
15:19 probably not a bad idea to outline it.
15:21 Have Chat GPT create an outline for you.
15:24 And then you can choose to do like like
15:26 there's certain things that make it more
15:28 problematic. If you want consistent
15:31 characters, that's harder, right? We're
15:33 going to play with that tonight.
15:36 Um dialogue and acting is harder. So you
15:41 could say I I want a silent film where
15:43 the story is just told visually. Um so
15:47 so get the get the the basic structure
15:50 right. Then you have to pick a style.
15:53 And so having something like my mind
15:55 where or or just go on to X and search
15:58 for
15:59 SRESFs, but you know, you have something
16:02 like this where you're just like, oh,
16:03 this is kind of a cool
16:06 style. Yeah, that's not bad,
16:10 right? That could be a film. It's sort
16:13 of a like a painting somewhere between a
16:15 painting and a
16:18 film. Or you could just, you know, you
16:22 could go into chat GPT if you're just
16:23 trying to find styles. You could go,
16:29 um, let's see.
16:32 [Music]
16:39 Um, I want you
16:42 to tell
16:45 me the top 20 films that you have that
16:49 you feel have the most
16:51 unique look and
16:54 feel that is
16:58 memorable and iconic.
17:04 period. Oh, and I'm I'm in 03, which is
17:07 the reasoning engine, so it's going to
17:08 think a little
17:18 bit. Completely new. Okay. So, so Jo's
17:22 digital completely new to AI. Okay,
17:24 great. Oh, okay. Cool.
17:27 So, Oh, completely new to AI. Okay. So,
17:30 go to chat GPT. Just start with chat
17:32 GPT. You don't need to do anything else.
17:34 And then you may have access to this
17:36 thing all called 03, but ignore it. I'm
17:38 going to I'm going to start
17:40 over. Oh, maybe I shouldn't. Should
17:45 I? Yeah. You know what I'm going to do?
17:47 I'm going to grab this list because this
17:49 is not a wasted list. 20 films that look
17:52 like nothing else.
17:54 Okay.
17:56 So, I'm going to copy that.
18:00 I'm gonna go nude. So here here we are
18:02 brand new. So first you should go to the
18:05 salon. Okay. So Joe goes digital. Anyone
18:07 else who's new here? Um the answer to
18:11 the question, how do I learn AI is right
18:14 now you
18:18 can't. And I know that sounds like an
18:20 [ __ ] move on my part, right? Like you
18:23 can't learn. I'm smart. I can learn it.
18:26 No, no, no. You don't get it. It's
18:28 changing like like what AI is and what
18:32 it makes possible is changing every two
18:35 weeks. Something new is possible and
18:37 it's and it's accelerating. So so we're
18:39 we're in a phase of generative AI which
18:42 which kicked off in 2022. Thank you very
18:45 much Winston. Appreciate that.
18:49 um when chat JPT launched that that that
18:52 we entered a new phase of AI which is
18:55 the phase where you don't have to be an
18:57 engineer to use AI right because chat
19:00 GPT made it a consumer app
19:02 effectively so we're not that far into
19:05 it so we're very early a lot of people
19:08 still haven't used it don't know what
19:09 chat GPT is okay so that's the start is
19:13 just accept that you can't learn it
19:16 because it's changing too fast. The
19:18 implication of that is get your ass in a
19:22 community. The AI salon is a community I
19:24 started the week after Chat GPT launched
19:27 as a gathering place for AI optimists
19:29 and creators. We meet twice a month and
19:33 there's a vibrant online community where
19:35 you can share and you know people teach
19:38 these things called learn out louds and
19:39 there's gathering spaces. There's sort
19:42 of this mastery area. Um, it's a it's a
19:46 remarkable community and there's lots of
19:47 other ones but but I'm fond to it. So go
19:50 to the salon.ai, click on join our
19:53 community and join. And the other thing
19:55 you should do is don't just lurk, join
19:58 join and introduce
20:00 yourself. And and if you're brand new to
20:03 this [ __ ] say that this is not a group.
20:05 There's there's a lot of communities out
20:07 there that are sort of like the
20:09 meritocracy thing where everyone's like,
20:11 you know, swinging their their big AI
20:13 wieners at each other. Those are
20:18 insufferable. The salon's not like that.
20:20 The salon's like, "Where are you? Let's
20:22 let's help you figure it out and get up
20:24 to speed and get you experimenting and
20:25 play it." All right, so that's the first
20:28 thing. And then just start
20:30 playing. You want to make a film. All
20:33 right. So, start out in chat GPT. One of
20:36 the first things I do is I say, "Okay,
20:39 let me let me give you uh a primer about
20:42 large language models in general. The
20:45 little prompt, we call it the prompt
20:47 hole, the the the chat hole, this little
20:50 box that you type in, it looks like a
20:53 lot like a Google search thing, but it's
20:55 nothing like a Google search thing." The
20:57 way to think about large language models
20:59 is you're starting a
21:06 conversation. Oh yeah, the the how can I
21:08 help you, Kyle? I don't know. This is a
21:10 relatively new thing. Probably within
21:12 the past two weeks, I've noticed that.
21:16 Um, treat it like a conversation. And
21:18 here's here's the the biggest thing you
21:20 need to know about large language models
21:23 is the more context you give them, the
21:26 better they get. So you can assign them
21:28 roles. I want you to act like a
21:30 screenwriter or you can you can say,
21:33 "Help me understand screenplay
21:35 structure." And it it will explain it to
21:37 you. And then it knows you're writing a
21:38 screenplay. And and now it has the the
21:40 context of that structure in its memory.
21:43 So every chat is kind of like this this
21:47 you're building like a Lego building of
21:50 conversation and and the more context
21:54 information you give it about what
21:56 you're trying to accomplish, the better
21:58 it gets. If you're really stingy with
22:00 your prompts, like write me a story that
22:02 I can turn into a film, it's going to
22:04 give you a shitty story for a shitty
22:07 film. You can go make it, but it'll be
22:08 shitty. Okay. So, here's here's and I've
22:11 written uh personally I've written seven
22:14 featurelength screenplays. My current
22:16 company is an automated video
22:18 storytelling platform. So, this is kind
22:20 of my jam. Okay.
22:22 So generally where I start is I'll just
22:25 start with
22:30 um I want to create a
22:36 compelling short
22:40 film.
22:43 Um and but let's see. But before we
22:50 do, help me
22:52 understand basic story
22:56 structure
22:57 using three
23:00 different
23:06 methodologies and then just let it do
23:08 its thing. Absolutely. So, let's break
23:11 that down. The hero's
23:14 journey. All right. It's going to tell
23:16 us it's it's a little slow tonight, but
23:21 whatever. Generally, I will do these
23:23 things and I will quickly glance at
23:25 them, but I won't read them. Thank you
23:26 for the chilies, Winston. Thank you so
23:28 much. Appreciate
23:30 that. I've been using chat GPT for five
23:33 months, so I understand what you're
23:35 saying. Others might benefit.
23:38 explain what different versions work
23:40 better for different things. Yeah, I'll
23:42 do that in a
23:44 bit. Um, but I just want to get the
23:47 basics. So bummed I miss missed Pate's
23:50 demystifying a this Easter weekend. I'm
23:52 sure he recorded it. Is it? It's the
23:54 salon.ai, Joe. Not just salon.ai. Can
23:57 you pop the URL back up
24:01 there,
24:03 Brandon? There you go. The salon.ai.
24:06 Sorry about that.
24:08 Okay. Um, so it told us about heroes
24:11 journey. It told us about three-act
24:13 structure for Hollywood. And then Dan
24:15 Harmon's story circle. Okay,
24:17 great. I often won't even read this
24:20 stuff, but now it has given itself a
24:24 bunch of context about we're going to,
24:26 you know, I want a well structured story
24:28 here, right? So now I'll say something
24:29 like
24:31 um let's
24:35 make it a
24:39 60-second one
24:42 character story
24:45 with story with no dialogue.
24:51 Um, give
24:52 [Music]
24:54 me
24:57 20 memorable ideas that are
25:03 either
25:06 heartbreaking or
25:11 uh,
25:14 humorous. All
25:15 right, we have a few starred comments.
25:18 Hang on. Let me go look at the start
25:19 comments in
25:24 StreamYard. Let's see. Kevin Stewart on
25:28 Tik Tok. I was at our AI conference for
25:31 the company this Friday. Awesome.
25:34 Fantastic. Keep getting educated. Make
25:37 keep Okay. If you're not making yourself
25:40 visible within your company as an AI
25:43 nerd, knock that [ __ ] off. Start geeking
25:46 out about AI and start talking about
25:49 it. In the in the salon, we call it
25:51 learning out loud or generously leading.
25:54 Like teach people what you're
25:55 learning. And if your response is
25:58 everybody thinks I'm stupid when I talk
26:00 about AI, let them let them think you're
26:02 stupid until they get laid off and
26:04 they're like, could you teach me that AI
26:06 thing? All
26:10 right. I just don't have tolerance for
26:12 these people anymore. I on China on Tik
26:14 Tok have an AI pitch for the pitch for
26:18 the VP tomorrow. That's
26:21 funny. Um, what's this one? Mr. on Tik
26:25 Tok. Today I turned 50 plus 13. Happy
26:28 birthday, Mr. It. Happy birthday, sir.
26:32 Fantastic. All right. And then what's
26:34 this
26:35 one? Valerie. My fear of AI is how it
26:38 will fit in our lives, but really it's
26:40 how we will fit in it. Yes, I agree with
26:43 that. To get the most of it means the
26:47 way we live today will have to change.
26:50 Many are not willing to do this change.
26:52 See, here's the
26:55 thing.
26:57 Um, do you
27:00 remember do you remember that
27:02 day? Yeah, we're going off on a rant
27:04 here. Brandon
27:06 knows the the whole people are not
27:09 willing to
27:10 change.
27:12 People don't want to
27:14 change. People claim they're not willing
27:17 to change. But do you remember that day,
27:19 I don't know, four or five years
27:21 ago
27:24 when all of the world governments
27:27 essentially on the same same day said
27:30 don't leave your
27:32 house and we
27:36 didn't. People are willing to change.
27:39 Now in that case it took the fear of you
27:43 know a worldwide you know sci-fi
27:47 contagion killing everyone to do it but
27:50 we did it literally
27:52 overnight. So, so people can absolutely
27:55 change. And then the the question the
27:57 question becomes this and I I think your
27:59 point's a really good one
28:01 that AI is getting so
28:05 quickly it's accelerate it's
28:08 accelerating sort of at our intelligence
28:11 right now and it's very quickly going to
28:12 be significantly above our
28:14 intelligence and there's a point at
28:16 which you give into that and you just
28:18 kind of accept it and say [ __ ] it just
28:20 like we gave into the steam engine and
28:23 machines that made us stronger.
28:26 Right. You know, there's a there there's
28:28 a point at which when the steam engine
28:30 came out that the the strongest man in
28:32 the village would have like a tugof-war
28:34 contest against the steam machine. And
28:37 there's a point at which the steam
28:38 machine just was too good. And and even
28:41 the strongest guy in the village was
28:42 like, "Yeah, [ __ ]
28:44 it." Right? That's what's going to
28:46 happen with AI. Uh Tik Tok pin. The
28:50 thing to fear about AI is the bad
28:52 assumptions it can
28:55 make and it's I don't know that AI makes
28:59 assumptions and its
29:01 overconfidence. I don't see AI as being
29:03 overconfident. I see AI as being overly
29:07 helpful which I guess is the same thing.
29:10 It will confidently here's one of the
29:13 ways I've explained large language
29:15 models chatbt in particular is
29:17 mansplaining as a
29:20 service where where a large language
29:23 model will confidently answer answer you
29:26 even if it's wrong and then even if you
29:28 point it out it will say oh you're
29:29 absolutely right I'm wrong and then
29:31 it'll be wrong again.
29:34 Um but it's what what was your point?
29:37 the mistakes. Wait, and if you're using
29:40 it for code, it will make mistakes that
29:42 will crash your development plan. Not
29:44 necessarily. Okay, listen. There there's
29:46 some of the comments in here are you
29:49 guys are veering very close to
29:52 tropes that you've heard pared about AI
29:56 that are very quickly becoming not true.
30:02 Coding with AI two years ago was
30:06 essentially unusable. Coding with AI now
30:09 is frustratingly maddening, but pretty
30:13 [ __ ] close. Coding with AI a year
30:16 from now is probably going to be better
30:18 than most coders out
30:21 there. Like like all of the major
30:26 um frontier model companies have said
30:28 that. So be careful not to just parrot
30:31 what people have said about what AI was
30:33 two or three years ago. Like you got to
30:35 be playing with this stuff. The new 03
30:37 model and the new 04 um mini and mini
30:41 high models from from chat GPT are
30:44 fundamentally different tools than
30:47 anything we've ever had
30:50 before. They're fundamentally different
30:53 tools. And I can explain why, but we're
30:55 in the middle of doing a a screenwriting
30:57 thing here.
30:58 So, so just be careful not to just
31:01 parrot what people have said. Here's
31:04 here's my my fear on your
31:08 behalf. If you just parrot what's wrong
31:10 with AI and and do that from the
31:14 sidelines, you're going to get caught
31:16 off guard and it's going to suck. So,
31:19 what I would say is go in with a good
31:21 healthy skepticism, but go in like get
31:24 on the [ __ ] field. Don't be up in the
31:26 stands with your arms crossed like ah
31:27 it's not really good at coding or
31:29 whatever. All right. Anyway, let's get
31:31 back to the story. Okay. So, he asked
31:33 for for 20 concepts. So, gave us 20
31:36 concepts. The last slice. A man prepares
31:39 for a romantic candle lit dinner. He
31:41 places two plates then eats one alone.
31:44 The quiet chair sits suddenly empty. Oh,
31:46 that's heartbreaking. All right. These
31:47 are good. Whatever. I'm I'm just going
31:49 to ignore these and then I'm going to
31:51 go.
31:52 Um, those are cliche and
31:59 expected and I
32:02 want
32:04 to win
32:07 awards for how good this story is.
32:13 Try
32:14 again and stop with the [ __ ]
32:24 emojis. You can yell at it. Some people
32:27 think you shouldn't because chatbt now
32:29 has memory, but
32:32 whatever. Did it Did it do it without
32:34 emojis? It did. Okay. A man carefully
32:38 lays items out on a table. hairbrush,
32:40 glasses, book. Throughout the
32:43 day, he sits beside each one and mimics
32:46 a conversation, revealing they're all
32:47 parts of himself he's trying to
32:49 remember. Jesus Christ, that's sad, but
32:51 that's got dialogue in it. Um, or maybe
32:54 it doesn't. An elderly woman catalogs
32:57 cassette tapes. Okay, these are all
33:00 sad. Anticipation, humorous, but not
33:04 silly, smart, unexpected. But a man
33:05 prepares to answer the door, fixes his
33:08 collar, adjusts the lighting, rearranges
33:10 books to look smarter. The buzzer never
33:12 rings. He sits fully staged alone, and
33:16 we realize it was never about the guest.
33:19 Huh. I wouldn't call that humorous. Uh,
33:23 okay. Whatever. It doesn't matter. Okay.
33:25 So, now we have some
33:28 stories. So, now I think I've still got
33:30 copied in my clipboard all those
33:32 different films. Now I want to
33:37 pick a
33:42 look
33:44 and feel and like
33:47 these films. And this was the list where
33:51 I said
33:53 um give me something visually
33:56 stunning. Did I Is did I Yeah, they're
33:59 all in
34:00 there. Oh, wait. Hang on.
34:03 Stop updated and save memory. Um
34:08 um tell me which
34:14 are the three that you
34:19 feel look the most
34:24 like a one
34:27 minute film. You can ask it to do stupid
34:30 [ __ ]
34:32 um of the kinds of stories you
34:38 wrote. That's a really stupid thing to
34:41 do, but why? Like, give it some
34:42 criteria. Barry Lyndon,
34:45 why? It's the gold standard for visual
34:48 poetry in stillness. Oh, I like
34:52 that. Great. The Red Shoes, 1948. Hero,
34:57 2002.
34:59 Each sequence hero is built around a
35:01 single color and an emotional truth. Oh,
35:03 I like that. Um, let's combine. Let's
35:07 combine one and
35:11 three. And I want you to name this new
35:17 visual
35:19 style. And I spelled that wrong. It
35:21 doesn't [ __ ] matter.
35:23 I want you to name this new visual style
35:27 and describe
35:30 it in
35:33 detail as if you were talking to a DP,
35:39 director of
35:42 photography. Mytho paint realism, aka
35:45 the Barry Lynon hero fusion. The style
35:48 isn't just about beauty. It's about
35:50 elevating emotional truth to myth.
35:53 using naturalistic lighting and da core
35:56 visual principles. Okay, so now it's
35:59 writing all that. So while it's writing,
36:01 I'm gonna I'm not going to even read
36:02 what it
36:03 says, right?
36:05 Because here's what we can
36:08 assume. It knows enough about those
36:11 films to understand them. It it
36:14 understands the story of those films. It
36:16 understands the cinematography of those
36:18 films, the visual cues.
36:22 it understands what stories it just
36:24 wrote. And one of the things about the
36:26 new chat GPT in the past since since
36:29 March
36:31 31st is that it's got a native image
36:34 model built into it. And what that means
36:37 is it can it can understand the language
36:40 of imagery and the imagery of language.
36:44 I know that sounds [ __ ] weird, but
36:47 you know, like we do, we can hear words
36:49 and picture something. We can see
36:51 pictures and make words, make stories,
36:54 right? Um, it's good at that now. All
36:56 right. So, there's all that. Great.
37:03 Um, I want you to
37:09 pick the one
37:13 uh
37:16 heartbreaking and spelled wrong and one
37:21 humorous story spelled
37:24 wrong. That you feel best
37:31 uh fit this visual style. So, so here's
37:37 an interesting thing. We're now picking
37:38 a story based on how it best fits a
37:41 visual style. With AI, you can really
37:44 start to mix the direction of how you do
37:46 things. Like normally, if you're going
37:49 to do a film, you write a screenplay
37:50 first. So, you got to do the writing
37:52 first. You rarely would say, "Here's my
37:54 story board." If you don't know what
37:56 [ __ ] story is, right? It's a story
37:58 board. But here we can start to mix
38:01 those things. So I got to a visual style
38:03 and now I'm going to say, "Well, you
38:04 wrote a bunch of stories. Which ones fit
38:06 best?" YouTube spotlight. Let's see.
38:09 Jeff Lan, I found it useful to have one
38:11 model in chat GPT. Fact check another.
38:13 That's great. They often catch
38:15 falsehoods that I didn't. That's a great
38:19 [Music]
38:21 idea. Just don't ask it to surprise you
38:24 when asking for a picture. Lol. It will
38:26 surprise you. That's very funny. Okay.
38:28 Heartbreaking pick. Still life. A man
38:31 paints the same bowl of fruit every
38:34 morning. Each painting slightly altered.
38:37 Pull back dozens on the wall. In the
38:40 center, a single photograph of his
38:42 partner's hands arranging the fruit
38:44 long, long ago. Oh my god, that's
38:47 heartbreaking. Source camp wants to go
38:50 to wants to go happy. What? Huh?
38:56 Oh, you want to go happy? Yeah, that
38:58 one's heartbreaking. All right, we can
38:59 go happy. Humorous pick. Anticipation. A
39:02 man prepares. If If we're not all
39:06 emotional if we're not all emotionally
39:08 prepared for the uh for the
39:10 heartbreaking film, then then let's go
39:12 happy. We'll go
39:15 humorous. My husband doesn't see images
39:18 in his head. I had Let's see. I had to
39:21 picture what that looks like. I Yeah,
39:24 that's weird. Um, can we use the video
39:27 chat GPT for Tik Tok as long as it stays
39:30 in the guidelines? Yeah. Yeah, you can.
39:34 I I'll show you that. We'll jump over to
39:36 Soro at some point here.
39:39 Okay. Uh, what's the story here? A man
39:41 prepares to answer the door, fixes his
39:43 collars. Uh, yeah, that's I don't quite
39:45 understand how that one works, but
39:46 that's okay. So now what I'm going to
39:48 say is I want
39:51 you to
39:54 determine the most
39:58 visually
40:01 poignant frame of the film
40:06 and mock that up for me.
40:13 Let's go vertical since I will
40:19 share on the
40:24 socials. All right. And so now we get to
40:26 see some of Chat GPT's image image
40:29 generation. Winston sent a paper crane
40:31 gift. Thank you very much, Winston.
40:33 You're very generous. I appreciate
40:35 that. Don't tease him with Sora. We
40:37 never go to Sora. Yeah. The the thing
40:39 about Sora is Sora kind of sucks. Sora
40:41 is long in the tooth. Um, but the new
40:44 image generation model inside Chat GBT
40:47 is is quite stunning. It's it's quite
40:54 powerful. Now, let me before this thing
40:57 gives us what it's making us right now,
41:00 there is a very high likelihood that
41:02 after all this work, the image that it
41:05 generates will just suck.
41:08 and it won't like it just won't light my
41:10 fire. Part of part of what you get to do
41:13 with
41:15 AI creative exploration
41:19 is move into a mindset of creative
41:22 abundance where you don't have to
41:24 [ __ ] it doesn't have to be perfect.
41:26 If you don't like it, start over. Like
41:29 it just doesn't matter. Wait, this did
41:31 the fruit
41:33 balls. This is the wrong This is the
41:35 wrong story. Did I tell
41:38 it? Oh, I didn't I didn't I didn't tell
41:41 it. We were doing the nonsad one. Oh my
41:44 god, it made it
41:45 [Laughter]
41:52 sad. Sora makes better images and makes
41:54 two at a time. Okay, great. Cool. We'll
41:56 hop over there. I was on Leonardo today
41:58 and got totally
41:59 discombobulated. It's all twisted around
42:01 now to me anyway. Yeah, I know.
42:03 Everything's changing.
42:06 Oh, there's the there's the wife doing
42:08 that. Uh, let me see.
42:13 Uh, plot change. Period. I forgot to
42:17 tell you that I wanted the most poignant
42:19 frame from the humorous
42:21 film, not the heartbreaking one. Period.
42:24 I'm too emotionally fragile for the
42:27 heartbreaking one. Period.
42:30 So, give me this one in a
42:34 style
42:37 that
42:42 [Music]
42:55 matches. What I just
42:57 put I just put its definition
43:00 of the style mythopoint realism.
43:03 Sometimes it kind of forgets like
43:05 because I'm jumping around a bit
43:07 sometimes it forgets what we were
43:08 talking about. So I just reminded it
43:10 what's what the style is again. Amos
43:13 still
43:13 [Laughter]
43:16 life. We'll say that the new image gen
43:18 is doing better at spelling. It's much
43:20 better at spelling. Much better. Uh it's
43:23 not perfect. Tik Tok comment. Um, I
43:26 don't see a Tik Tok
43:28 comment. Joe goes digital. I'm thinking
43:30 about using short films to make ads for
43:32 my digital marketing. Cool. Yeah, if
43:35 you're doing them for marketing, then I
43:37 would say instead of a a story framework
43:40 like I did, ask for a marketing plan and
43:42 then ask for a social media plan and
43:44 then tell them what you're you can point
43:47 your
43:48 um just put in your business URL and say
43:52 I want a marketing plan for this
43:53 business that focuses around short form
43:56 video storytelling intended for Tik Tok
43:59 and it'll come up with
44:02 it. Wait, what?
44:06 That's not from the story you gave
44:09 [Laughter]
44:16 us. The
44:18 humorous story
44:24 is make the
44:26 frame from that
44:30 film you buttthead.
44:34 [Laughter]
44:43 Never a dull
44:51 [Music]
44:57 moment. He's not moping around. He's
45:00 moping around. I know. That was really
45:02 [ __ ]
45:09 funny. This looks like a
45:12 painting. Okay,
45:15 listen. This is also sadder than
45:19 [ __ ] Okay. Okay.
45:22 Um, this
45:25 looks like a painting or
45:30 cartoon, not a
45:33 film. Make it
45:36 photographic and make
45:38 it not so
45:41 flipping
45:46 depressing. No YouTube. Yeah, we're live
45:49 on YouTube. I went live, didn't
45:52 I? Yeah, we are. We are live on
45:56 YouTube. Sea slug of
46:00 doom. If you're doing marketing, build a
46:03 brand bible and get results from chat
46:05 GPT. Yep, I would agree with that. And
46:07 any AI is on brand. Yeah, you you can
46:10 basically take your brand bible and
46:12 upload. You can either make a custom GPT
46:15 or just upload it, you know, into a chat
46:17 session.
46:21 Oh,
46:23 okay. I just made a website using chat
46:26 GBT. Yeah, it's remarkable, isn't
46:28 it? That still doesn't look very cinem
46:33 photographic. That's a little better.
46:35 All right. It's a little
46:37 [Laughter]
46:40 better. All right.
46:42 Now, so now what you can do is now chat
46:45 GPT kind of knows what the general
46:49 um look and feel is. You can say,
46:53 um, give
46:55 me
46:57 a
47:00 description of the shot
47:04 byshot
47:09 storyboard. And I'm not going to I'm not
47:12 going to continue on with the entire
47:14 process here
47:16 because you're doing something slightly
47:18 different. But I think you get the
47:20 point. The point is lots and lots of
47:22 context, lots of lots of ideiation and
47:26 then you can take the best of what it
47:28 generates and like if it doesn't
47:29 generate anything good, tell it. Tell it
47:32 why it didn't, you know, why it's not
47:34 good. Like what is it about what it gave
47:36 you that's not what you wanted? And then
47:38 in articulating that, it will start to
47:40 focus it into what you want. And then at
47:43 some point you can just sort of take all
47:44 that [ __ ] and put it into a new chat and
47:48 start over,
47:50 right? All
47:55 right. 60 seconds. Shot one preparation
48:02 [Music]
48:11 begins mirror
48:20 monologue. So let's
48:24 see. Generally what I'll do here is I'll
48:27 take these shots. I'll
48:30 say
48:31 make this shot in our
48:36 style. So I'll just copy and paste what
48:39 it just told me because sometimes you
48:42 think it it it can see what it what it
48:47 wrote and then sometimes it just
48:49 doesn't. So, I just find it a little
48:51 less frustrating to just copy and paste
48:54 [ __ ] like
48:59 that. I love the give me 20 fill-in-
49:02 thelank method. I also learned it here.
49:11 Nice. 20 bucks a month for Chat GPT. You
49:15 can do most of what Chat GPT does in the
49:18 free plan. Uh, but especially if you
49:20 want it to be
49:24 um if you want to not get rate limited
49:26 out, I'd pay the 20 bucks a
49:29 month. Oh, this is cool. That's actually
49:32 a really cool shot. Oh, that's really
49:34 cool. Let's let's do let's do three of
49:37 these. So, we'll go grab shot
49:45 two. Then I'm going to go copy shot
49:58 three. Kyle, they are giving me
50:00 subscription for my birthday. What
50:03 should I choose? Oh, that's
50:06 sweet. Um, I don't know what you want to
50:08 do.
50:11 I mean, I would say choose a
50:13 subscription for something that you're
50:15 frustrated you can't do enough
50:28 of. All right. How's this
50:40 [Music]
50:42 coming? I hate being timed out,
50:45 especially when I'm having a deep
50:46 conversation. Oh, it's [ __ ]
50:48 maddening. [ __ ] maddening.
50:58 This is the most depressing, humorous
51:00 story I've ever
51:02 seen. It's a guy getting ready for
51:04 someone to come to the door and they
51:06 never
51:09 come. He's laughing on the inside. Yeah,
51:13 exactly. It's it's it's sad. All right,
51:16 there's number two. Let's see. We done.
51:19 And then we'll do number three. And then
51:21 while that's doing that, what I'm going
51:23 to do is I'm going to copy this and then
51:25 we're going to head over not well maybe
51:27 we should head to
51:28 Sunno just to because or Sora because I
51:31 haven't played with Sora in for [ __ ]
51:34 ever. So maybe it's a little better. So
51:36 we're going to go to
51:37 videos and I should be able to paste.
51:41 Yeah. So, I just pasted in him doing his
51:44 thing. And I'm going to say, "Man
51:48 buttons his
51:51 shirt
51:53 carefully. We
51:56 never see his
52:00 face." And then I'm going to see if I
52:02 can switch to I can only do 2x3. Oh, but
52:05 wait, let's go video. And then I can do
52:08 9 by6.
52:11 So, we'll do that. 5 seconds is fine.
52:16 And we'll do 480p is fine. All right.
52:18 Here we go. Let's see. Do we want to go?
52:21 Yeah, there we go.
52:24 Generate. Added to the
52:27 queue. My
52:29 media. All right. So, that's rendering.
52:32 There's me driving a
52:34 truck. Go Speed Racer. Okay. Then we'll
52:39 go back here to this thing. We'll get
52:42 number
52:43 two. Copy image. Go back to Sora. Paste
52:48 it
52:49 in. And then we're going to do video
52:51 again. And we're going to
52:54 say
52:56 man
52:58 carefully
53:01 arranges the
53:04 picture on the table.
53:12 All
53:19 right. All right. We got two versions of
53:21 our
53:29 video. That was
53:31 [Laughter]
53:34 weird. All right.
53:43 Yeah, those are both
53:47 unusable. It's like he's putting on a
53:51 microphone. So, let's
53:55 take let's take this one and go over
53:59 to I guess Luma
54:02 Labs. Lumalabs should be better.
54:06 So I'm going to paste this
54:08 in and say
54:12 man
54:16 carefully
54:20 finishes buttoning
54:23 his
54:29 shirt and lowers his
54:32 hands. I don't know. It It generally
54:35 doesn't follow directions like that for
54:36 some weird ass
54:39 reason. Um 720p I think going to match
54:43 the
54:46 um aspect ratio that we uploaded. So it
54:50 won't quite be 9 by6, but we'll take a
54:52 look here. All right. Gerlin, right? Not
54:55 impressed with the dancing
54:58 cat. Uh um do cows getting ab abducted
55:03 by aliens. See if Sora can do that. He's
55:05 wearing a wire. It totally looked like
55:07 he was putting on a wire in that Sora
55:09 thing. All right, let's go look at one
55:10 of the other Sora ones.
55:15 Um let's
55:18 see. Here we go.
55:22 [Laughter]
55:28 The couch has glass on
55:31 it. That's so bad. That's so
55:37 bad. He just changed
55:41 jackets. Yeah. Sort of sort of
55:50 sad. Okay. So now that's
55:54 interesting. We got bad character
55:57 consistency. So I'm going to copy this
55:58 image. I'm going to paste it into chat
56:00 GBT and say that
56:04 third
56:06 image is a
56:09 completely different person. You
56:15 ninny. Um, use this
56:19 as character
56:25 [Music]
56:29 reference. We'll see if it fixes
56:35 it. And then what I want to do is I want
56:38 to go over to that to that uh Twitter
56:42 post about
56:46 um the character consistency and go
56:49 replicate that because there's a bunch
56:51 of cool things we can do with this
56:57 post. But let's go let's go see if how
57:01 our image is doing. Let's go back to
57:03 Lumal Lab. See if that's things done.
57:05 Nope. Waiting your
57:08 turn. Those are completely useless. This
57:12 is the thing with Sora is Sora for some
57:15 things is quite good and then but for
57:17 other things it's just
57:19 like I I don't know what they trained it
57:21 on. It's It's weird. It's weird, man.
57:25 You know, man, it's so weird,
57:28 man. Dang it. All right. PNG. Yep. Yep.
57:33 Yep. Yep. Yep. Yep. Yep. Yep. Yep. Yep.
57:36 Yep. Yep. Yep. All right.
57:40 How we doing on likes over here on Tik
57:42 Tok? Do me a favor. Start doing some
57:44 little tappy taps on Tik Tok. Let's see
57:46 if we can get up. We're We're at 8,000.
57:47 Let's see if we can get to like 25,000
57:49 shortly. All right. Did it Did it give
57:51 us a guy in the same Now he's not
57:53 wearing a
57:55 jacket and he's wearing a blue
57:59 shirt. And here it's a white
58:02 shirt with a
58:04 jacket. Well, whatever.
58:13 request from Tik Tok live. What is it?
58:15 What
58:20 games? All right. I don't want to I
58:23 don't want to run down that rabbit hole.
58:24 That'll distract
58:27 me. Almost wearing a
58:33 smile. Yeah, I can't make this film
58:35 anymore. It's depressing the [ __ ] out of
58:38 me. He looks like a young Tom
58:42 Hanks. Um, okay. Let's
58:46 see. Let's go to Lumal Labs. They do
58:48 their thing yet? Nope. Come
58:51 on. Waiting your turn. This is one of
58:54 the one of the bad things about Lumalabs
58:56 right now is their servers are so hosed
58:58 because they've they're they're one of
58:59 the good ones and so it's really hard to
59:02 get something that works.
59:04 Okay. So, what we've got on screen right
59:07 now is this post that said several weeks
59:10 ago, and generally I don't like these
59:12 things that are like, you know, here are
59:13 the 17 steps for for doing something
59:16 ultra specific, but this is one that
59:18 I've been meaning to do anyway. I want
59:20 to do this for my musical Sydney is do
59:23 character sheets. And he just the way he
59:25 did these is really smart. So, he goes,
59:28 "Several weeks ago, I decided to test
59:29 chat GPT40's character consistency
59:32 abilities. I must say I was pleasantly
59:34 surprised. Here's how I did
59:37 it. So, you pop in there and then what's
59:40 cool is he's given us screenshots of
59:43 what he did in chat GPT. Now, he gave us
59:46 the the the prompt over here as well,
59:50 but we could also just, you know, take
59:53 this screenshot and have chat GPT read
59:56 it. Let's see. I want to m maintain a
59:59 thread here. The idea is to develop an
1:00:00 understanding
1:00:03 All
1:00:07 right. Oh, he pulled a curated set of
1:00:10 images from
1:00:12 Midjourney, but they weren't all from
1:00:14 the same genre, but they shared a
1:00:17 tone. All
1:00:28 right. So, what did he say?
1:00:32 Perfect. Okay, I
1:00:34 see. All right. So, let's go to
1:00:36 midjourney. Let's Let's just do what?
1:00:38 Let's We'll follow his
1:00:40 instructions. Anything here?
1:00:43 No. Oh, midjourney. Midjourney, where
1:00:46 are you? Where are midjourney?
1:00:50 Midjourney. Mid Journey. Rather, carry
1:00:52 on. Cheerio pip pipep. All right, let's
1:00:55 go to
1:00:57 organize and then
1:01:03 goank. Look at
1:01:08 [Laughter]
1:01:11 doggies. Oh, this would be cool. We
1:01:14 could get Yeah, let's do this style.
1:01:17 Okay, so let me grab this copy image.
1:01:23 Then I'm going to go back over to chat
1:01:25 GPT. We're going to go
1:01:27 new. We're going to go
1:01:30 bang. Going to go to midjourney. Oh,
1:01:33 what happened? Oh, here we go. And then
1:01:35 we're going to
1:01:36 go. So, which one? We did that one.
1:01:39 We'll grab this
1:01:44 one. Boom. I think you can put up to
1:01:48 five images in something like this.
1:01:53 Let's do this one. Copy
1:01:57 image chat.
1:02:00 GPT.
1:02:05 Boom. I kind of like that one's all
1:02:07 blown out. Kind of cool. Copy
1:02:09 image. Go back to chat. JP. Boom. We'll
1:02:13 do five of
1:02:14 these. So, one more.
1:02:20 Uh, I like that. Let's
1:02:34 see paint pin. New video model was
1:02:37 released today. Semi-interesting just
1:02:39 because it's auto reggressive. New new
1:02:42 video model where Pate in AI studio on
1:02:45 Google or a new video model
1:02:48 altogether from a from a different
1:02:53 source. Magi Magi one is that was that a
1:02:59 model was that an open source model that
1:03:00 was released today that was really
1:03:02 small? I think that was a voice model
1:03:03 that I I read
1:03:06 about. Let's do this one.
1:03:11 All
1:03:11 right, we'll pop that in there. Magi
1:03:14 one. Okay, cool. Um, I'll check that out
1:03:16 before tomorrow. Let's see. Okay, I
1:03:20 [Music]
1:03:22 want to How did he put it? Oh, let's
1:03:25 just go to his thing and we'll
1:03:27 screenshot
1:03:30 it. So, I'm going to
1:03:33 screenshot. I want to maintain a thread
1:03:36 here. The idea is to develop an
1:03:37 understanding of my approach to lighting
1:03:39 composition and color grading. Here's an
1:03:42 example. All right, that's
1:03:48 fine. We'll just take his prompt al
1:03:51 together. Okay.
1:03:53 Boom. And then we're going to drop that
1:03:55 into chat GPT as
1:03:57 well. I'm just going to say follow the
1:04:03 text on
1:04:05 the image as a
1:04:09 prompt about the other
1:04:15 images. We'll see how well it does this
1:04:19 out. Let's dive into starting to build a
1:04:22 thread of visual language, lighting,
1:04:23 composition, and color grading.
1:04:25 Beautiful visual thread analysis. Flash
1:04:29 heavy, overexposed aesthetic, nightclub
1:04:31 and ambient
1:04:32 mood. Overwhighted skin tones. In some
1:04:35 shots, the skin tones are lighter of
1:04:37 lighter skinned individuals are almost
1:04:40 ghostly composition casual framing. So,
1:04:43 this is cool. All right. So, it's it's
1:04:45 got all this stuff.
1:04:47 Great. So, let's see what he did next.
1:04:50 See, here's the problem with ADD. I'll
1:04:53 go to something like this where he's
1:04:54 like, "Here's my step-by-step process."
1:04:56 I get to step one and I'm like, "Oh, I
1:04:58 have three ideas. Let me run down a
1:05:00 rabbit hole." And I never finish the
1:05:01 [ __ ] project. Tik Tok question. How
1:05:04 good is the Dream Machine iOS app from
1:05:05 Luma? Pate. Uh, just new
1:05:13 model. Oh, new model at
1:05:17 sandbox.ai or sandai. Okay, got it. Open
1:05:20 source. Okay, cool. Yeah, I'll check
1:05:22 that out.
1:05:26 Um, so dream. So,
1:05:29 let when you go to Luma
1:05:33 Labs, they have this thing called um
1:05:36 what are these called? Boards. And if
1:05:39 you go into a
1:05:42 board, you've got this you you can
1:05:45 scroll up and down. And every one of
1:05:48 these is a prompt
1:05:51 uh with with either an image or a video.
1:05:53 And then for any one of these, you can
1:05:56 create variations of them and you scroll
1:06:00 horizontally. When I first saw this on
1:06:02 desktop, I'm like, this is a really
1:06:04 weird [ __ ]
1:06:05 interface. And then I used it on mobile.
1:06:08 And when you use this on mobile, it's
1:06:10 really intuitive. It's very much like
1:06:13 slidy slidy up and downy, you know?
1:06:18 um intuitive, you know, slide left and
1:06:21 right and then the way it the way it
1:06:23 writes prompts, it turns them into Mad
1:06:26 Libs where you can just do pull downs
1:06:28 and when you do a pull down, it
1:06:30 generates a whole new set of
1:06:31 images. Uh off-topic Tik Tok pin Pate
1:06:35 Mo, did you see the model
1:06:37 ECGMLP open source cancer detection?
1:06:41 Wow. With accuracy of greater than 97%.
1:06:45 Yeah, it Oh, by the way, if you haven't
1:06:48 seen it, there's a I don't I think it
1:06:50 might be Netflix and Brandon, maybe you
1:06:52 can find it, but there's a there's a
1:06:54 documentary about Demisabus and Deep
1:06:57 Mind, and it's really cool. It starts
1:07:00 with when they started Deep Mind. So,
1:07:02 they've got early footage of of Demis,
1:07:05 you know, at uh where wherever he was,
1:07:08 and then he went to Cambridge, and
1:07:09 they've got that's where he met his
1:07:11 co-founder. And um it's a really good
1:07:14 documentary. It's a little boring
1:07:16 because it's just him talking about how
1:07:18 he wants to invent AI and then he
1:07:25 does. He's like, "We need more compute
1:07:28 power. Oh, Google bought us. Hey, we
1:07:30 should do this. Protein folding seems
1:07:32 hard. Let's go do protein folding. And
1:07:34 the first year they like did really good
1:07:36 and the second year they did [ __ ]
1:07:38 awesome and won the Nobel Prize in
1:07:41 chemistry. So yeah, it's called the
1:07:43 thinking game. That's it. Um, but it's
1:07:45 worth watching if you want to if if you
1:07:47 want to
1:07:48 have a nice sense of optimism and if you
1:07:51 also want to be inspired by
1:07:55 someone from the time he was a little
1:07:59 kid, he knew he wanted to do what he's
1:08:02 doing now. There's something really
1:08:03 inspiring about that. He was also like a
1:08:05 chess champion. Like in his spare time,
1:08:08 he was a chess champion, but he was
1:08:09 always thinking about AI. It's streaming
1:08:12 everywhere. Okay, cool. Amazon, YouTube.
1:08:14 Great.
1:08:15 Beautiful. Kyle's tab groups need meta
1:08:19 meta grouping. They do. Uh, embrace the
1:08:22 jank still. Yeah, still embrace the
1:08:25 jank. Um, okay. Where were we? Oh, we're
1:08:28 back at this. I was going to go read a
1:08:30 second instruction.
1:08:33 Okay. Okay. So, it ended up with this.
1:08:36 Okay. First, we came up with a core
1:08:39 style module. Wait. So it says, "Please
1:08:42 provide the core style module for our
1:08:45 prompt." Oh, cool. I like this. Okay.
1:08:49 Five. Please provide the core style
1:08:52 module for our prompt in which our
1:08:56 character Lancelot is looking straight
1:08:58 into the camera with a neutral
1:09:01 expression. All
1:09:05 right. Got it. So, we're
1:09:08 gonna we're gonna copy that. We're going
1:09:10 to go to chat
1:09:12 japet. We're going to paste that and say
1:09:15 um
1:09:18 uh re write this for
1:09:23 me, which is just OCR it. So, I could
1:09:27 just retype that in. But here, here's
1:09:29 this thing. Or if you want a more
1:09:31 cinematic edge. Oh, good. It rewrote it
1:09:33 for us to make it more cinem cinematic.
1:09:36 All right, cool. So, we're going to take
1:09:37 this. This is now going to become our
1:09:39 prompt. We're going to edit it. Create a
1:09:41 foundation style for our prompt
1:09:42 character for for for our character Bob.
1:09:50 I'm going to say
1:09:52 Bob is an
1:09:56 aging game show
1:10:00 host who
1:10:02 somehow ended
1:10:06 up
1:10:08 working for a Tik Tok
1:10:17 creator who he hates.
1:10:19 [Laughter]
1:10:22 captured head head on gazing directly
1:10:24 into the camera with a calm unreadable
1:10:28 expression. That's interesting that he
1:10:30 wrote it that way. All right,
1:10:32 so let's see what this writes for
1:10:36 us. Prompt: Create a photographic
1:10:39 portrait of Bob, an aging blah blah blah
1:10:41 blah blah. Okay. style, harsh flash
1:10:44 photography, 80s magazine shoot
1:10:46 aesthetic, background slightly out of
1:10:48 focus, cheap LED ring light reflections,
1:10:51 Tik Tok prompts subtly subtly cluttering
1:10:56 the edge of the frame. Um, color grade
1:10:59 leans faded pastel with exaggerated reds
1:11:02 and purples. Composition tight
1:11:05 midshot. All right, cool. All right, so
1:11:08 now what's he tell us to do? I assume go
1:11:11 make the pitch. Lastly, the pose
1:11:13 variations. This is where I thought it
1:11:15 was really cool.
1:11:17 Okay.
1:11:20 So, to build consistent character from
1:11:23 all sides, I rendered eight distinct
1:11:26 body positions, turning the subject. So,
1:11:30 wait, I'm confused.
1:11:53 Okay. Okay. So, this is three different.
1:11:56 Okay. So, now Okay. So, what we're going
1:11:58 to do, we're going to copy this one and
1:12:01 we're going to go back to chat jetat and
1:12:03 say, I want you to make a character
1:12:10 sheet based
1:12:13 on the
1:12:17 attached tik tok pen. Kyle prompts are
1:12:21 like the writer's room on whose line is
1:12:23 it anyway? Yeah, you know, it is it like
1:12:28 there's a lot of folks are right now
1:12:31 talking about vibe coding. I've talked
1:12:34 about vibe
1:12:35 storytelling. I I think I think where we
1:12:39 go with
1:12:40 AI
1:12:42 increasingly is that we just get to
1:12:45 vibe and and it okay
1:12:50 listen especially if you're new here
1:12:52 especially if you're new to
1:12:57 AI there
1:13:00 is we talk a lot in here about having
1:13:02 your Kevin Mallister moment remember
1:13:04 from Home Alone that moment moment where
1:13:07 you're in this sort of moment of shock
1:13:08 and
1:13:09 awe. There's a lot of that with AI
1:13:12 that you go into it and no matter what
1:13:15 your expectations are, somehow AI is is
1:13:19 is going to is going to do something for
1:13:22 you
1:13:24 that you didn't think was possible. And
1:13:27 it's going to feel like science fiction
1:13:29 and it's going to blow your mind and it
1:13:30 might scare the living [ __ ] out of you.
1:13:33 Because the thing that it will that the
1:13:35 thing that will freak you out the most
1:13:37 is the thing that you had yourself
1:13:39 convinced AI couldn't really do that.
1:13:42 That's my thing. AI can't really do
1:13:44 that. And there's going to be a point at
1:13:46 which AI is going to do something that's
1:13:48 just going to [ __ ] freak you out.
1:13:53 But feel those feelings because it's
1:13:56 intense. It's weird. like especially if
1:13:59 you've been doing something for a while
1:14:01 and you're like it shouldn't be this
1:14:02 good and then it is. It's kind of a
1:14:05 trip. Um
1:14:08 but there is something incredibly
1:14:12 liberating about having a tool that does
1:14:15 does the drudgery work of like
1:14:18 execution. And what we get to do is kind
1:14:21 of, you know, skim along the surface of
1:14:26 an idea and just go, "Oh, give me a
1:14:28 picture. Give me words. Give me a
1:14:29 structure. Give me a framework. Give me
1:14:31 a this. Give me a marketing plan. Give
1:14:33 me a copy for an ad. Oh, make the ad.
1:14:35 Let me see what it looks like." It's
1:14:37 It's bonkers. It's kind of
1:14:40 bonkers. All right. Is this Would you
1:14:43 like me to generate the visuals?
1:14:45 Yes. Yes.
1:14:49 Do I look like
1:14:51 I'm
1:14:54 [Laughter]
1:14:58 patient? I've had that happen. Yeah, the
1:15:01 the ultimate mind mindset coach is AI. I
1:15:05 wanted to organize my accounts. Yeah,
1:15:07 the organization stuff with AI right now
1:15:09 is a bit of a mess. It's it's not super
1:15:12 organized. Um, all right. So, now I'm
1:15:15 going to go back to this thing. I'm
1:15:16 going to get these two other I'm going
1:15:18 to get this one
1:15:20 now. And then I'm going to go back to
1:15:22 chat GPT. I'm going to paste it in. I'm
1:15:25 going to say now make a
1:15:29 care
1:15:32 career
1:15:34 sheet based on the
1:15:38 attached and go right
1:15:42 into creating it.
1:15:48 Oh, this is going to be good. Look at
1:15:50 this
1:15:52 dude. This is going to be really
1:15:58 good. That That totally totally looks
1:16:02 like a bitter exame show host, doesn't
1:16:04 it?
1:16:14 It's got ring lights. Beautiful. I don't
1:16:17 know if it did all of our shots, but it
1:16:19 does matter. The price is wrong. Um, all
1:16:21 right. So, now we're going to do the
1:16:23 next one. Now, we're going to go back
1:16:24 over here. We're going to grab
1:16:28 this copy image. Then, we're going to go
1:16:31 back over
1:16:32 here and go now make
1:16:57 Oh my
1:17:05 god. Yes,
1:17:08 please. I'm so tired of chat GBT
1:17:12 deferring to me. Just [ __ ] do the
1:17:15 [ __ ]
1:17:16 I have ADD. I don't have to concentrate
1:17:20 and plan. I don't have to plan your
1:17:23 stupid [ __ ] I'm not a project manager.
1:17:26 I'm an idiot with
1:17:30 superpowers. He won a photo shoot. Oh
1:17:34 [ __ ] What's going on over here? Good
1:17:40 lord. Put that on a t-shirt. What did I
1:17:43 say?
1:17:44 [Laughter]
1:17:46 I'm an idiot with
1:17:49 superpowers. Actually, that's not
1:17:52 bad. Here, wait. We're going to do We're
1:17:55 going to do a new one. Let's go to a new
1:17:56 chat GPT.
1:17:58 Um, I want to
1:18:01 create an 80s
1:18:07 style
1:18:11 t-shirt with
1:18:14 big
1:18:16 humorous spelled wrong
1:18:19 letters. Um that
1:18:23 say in large
1:18:28 letters
1:18:30 AI and then below it in smaller letters.
1:18:48 Um,
1:18:51 it's like being an
1:18:56 idiot with superpowers. It's s
1:18:59 superpowers is two words,
1:19:02 [Laughter]
1:19:08 right? I think it would be funny funnier
1:19:10 if it just said I'm an idiot with
1:19:12 superpowers. But I like that doesn't
1:19:14 make any
1:19:20 sense. Looks like Stan Lee without the
1:19:22 stash. That's pretty funny. The rice is
1:19:25 right. The the rice is bright. AI Doomer
1:19:29 recently told me I was spreading misinfo
1:19:31 by saying prompt quality affects output
1:19:33 quality.
1:19:34 Man, that's a [ __ ] stretch
1:19:37 for What's the What's the risk there if
1:19:41 you're right in his
1:19:43 mind that these things could actually
1:19:46 potentially do good give good answers if
1:19:50 you stumble onto the right
1:19:53 prompt? Doomers are weird, man.
1:19:57 You you know what's you know what's
1:19:58 strange is
1:20:00 um with a lot of the doomers the tropes
1:20:03 haven't their their tropes haven't
1:20:05 evolved like their tropes are like it
1:20:08 hallucinates it steals it copy and paste
1:20:11 it it's going to kill us it's uh it's
1:20:15 environmentally [ __ ] up
1:20:17 like
1:20:19 just almost all of the tropes are the
1:20:21 same tropes they've been
1:20:24 saying Um,
1:20:30 uh, yes. Yeah. Uh, let's see. Yes.
1:20:34 Give me the
1:20:38 t-shirt
1:20:40 graphic in
1:20:44 incredible 80s design and make the
1:20:51 graphic with a
1:20:55 transparent
1:20:56 background. Did you know chat GBT image
1:20:59 gen can make transparent backgrounds
1:21:01 now? Bet you didn't know that. Or if you
1:21:04 did, you probably hang out here and
1:21:07 you're a [ __ ] weirdo. We're not
1:21:09 weird, Kyle. We're irregulars. I thought
1:21:12 we established that years
1:21:14 ago.
1:21:16 Maybe. Why are you so
1:21:19 defensive? Okay. With transparent
1:21:22 background. Boom. All right. Tik Tok
1:21:25 pin. Joe goes digital. I used to think
1:21:27 AI was evil, but I've come to terms with
1:21:28 AI is here to help, not the other. Yeah.
1:21:31 So Joe, Joe Joe, I first of all, I
1:21:34 appreciate you being here. I appreciate
1:21:36 you hanging out. I appreciate you asking
1:21:38 really good questions and I really
1:21:40 appreciate you admitting that I my
1:21:43 experience has consistently been that no
1:21:46 matter how hard the person was resisting
1:21:49 AI, when they actually experience it,
1:21:52 like what one of the first things I
1:21:54 experienced was, huh, this feels oddly
1:21:58 human, right? It wasn't, you know, well,
1:22:02 actually, one of the very first thoughts
1:22:03 I had in my head was, wait a minute, did
1:22:05 the Turing test just get passed? And the
1:22:07 answer is yes. Um, and in fact, I think
1:22:10 it was actually just officially
1:22:12 declared, but then like everyone keeps
1:22:13 moving the goalpost. Doesn't [ __ ]
1:22:15 matter.
1:22:19 Um, what I've come to learn about AI is
1:22:26 the concerns
1:22:28 about the data that was used to train
1:22:31 these models is absolutely valid. Like
1:22:36 the the Frontier model companies
1:22:38 absolutely went out and scraped
1:22:39 everything that they could get their
1:22:41 hands on. And I promise you, they used
1:22:44 nefarious scraping tools from salacious
1:22:48 countries all over the world that they
1:22:50 shouldn't have. Promise you they've done
1:22:53 that. The result is that we now have
1:22:56 this thing that is instant
1:23:01 access to like a collaborative
1:23:04 conversation with every human being
1:23:06 that's come before us.
1:23:08 And what it ends up being
1:23:11 is like you have an idea, you Joe have
1:23:15 an idea and you put it into this stupid
1:23:18 [ __ ] box and it like instantly
1:23:21 materializes that idea and amplifies it
1:23:24 like like we just made a t-shirt design,
1:23:26 right? It's like being an idiot with
1:23:31 superpowers.
1:23:33 Uh, awesome. Now make a '9s
1:23:41 design. Let's see what that looks
1:23:43 like. Like that was a stupid idea. Like
1:23:46 I said something stupid. Brandon said
1:23:49 that should be on a t-shirt. 3 minutes
1:23:51 later that idea is out of my head and in
1:23:53 the
1:23:55 world. That's like a humanity amplifier,
1:23:58 right? It's not killing the humans. It's
1:24:02 giving us the ability to to
1:24:07 to accelerate who we are and the impact
1:24:09 we want to have. And it and it's
1:24:12 incredibly inspiring. And so what I
1:24:14 found consistently is people that were
1:24:16 the hardest resistors tend to be the
1:24:18 biggest evangelists for it because they
1:24:21 they experience the gap between what
1:24:23 they thought it was and what it actually
1:24:25 is. So thank you for being here. Source
1:24:28 camp. Met a client today who started out
1:24:30 saying this can't help us to here are
1:24:33 the six projects we could do. Yeah,
1:24:37 exactly. Well, no, I I know about a I've
1:24:40 used chat GPT. It can't help me. I
1:24:42 tried. Yeah, I tried. What did you put
1:24:44 in? Uh, write me a recipe and it was a
1:24:47 bad recipe. That That was it. That was
1:24:49 the sum total of your experience with
1:24:50 ChatGpt. Oh, no. I use it all the time.
1:24:53 What do you use it for?
1:24:57 And then you show them what it can do
1:24:59 and they're like, "Oh my god, it can do
1:25:03 that."
1:25:04 Right. This is the third time I've tried
1:25:07 to scroll down.
1:25:10 Oh, this is a good one. I like the I
1:25:13 like the 90s one. Uh, let's do let's do
1:25:17 um now Matrix style.
1:25:27 And and this is these are these are
1:25:30 transparent. Like you notice when I roll
1:25:32 over it, it's got the checkerboard
1:25:34 background. That's real. These are
1:25:36 transparent transparent PGs. I could I
1:25:39 could either just take these as they are
1:25:40 or I could just upscale them in some
1:25:42 upscaler, right? if you're if you're an
1:25:45 insane OCDer about pixel or you know PPI
1:25:50 count. But quite frankly, nobody looks
1:25:52 at t-shirts that
1:25:55 close, you
1:25:58 know. Uh wait, we got to go back to our
1:26:01 dude. Dude.
1:26:16 [Music]
1:26:17 It chopped that off, but that's
1:26:19 okay. Now, we'll do this
1:26:22 one. Okay. So, here's here's how I would
1:26:25 use these these character sheets. So, we
1:26:28 have this character, right?
1:26:31 Bob, and we've
1:26:34 got this one. Didn't I have another one?
1:26:37 I thought I had another
1:26:39 one. Did I lose it?
1:26:44 Oh, here it
1:26:45 is. So, here's
1:26:49 Bob and here's
1:26:53 Bob and then we're going to have a new
1:26:55 Bob. So, we're going to have three
1:26:57 character
1:26:59 sheets. So, what I would do is I would
1:27:03 say I would take a story or a script. I
1:27:06 would have it break a scene down into
1:27:08 shots.
1:27:09 And then I would have these sets of
1:27:11 character sheets and I say, "Okay, this
1:27:13 is a scene between Bob and
1:27:15 Kyle." And I would just paste those
1:27:18 character sheets into the prompt and
1:27:20 just say Bob confronts Kyle and maybe
1:27:22 put what they're wearing,
1:27:25 right? And now we're going to get
1:27:27 consistent characters. And what's cool
1:27:28 is like you can start out like even
1:27:30 though these are photographic, you can
1:27:32 do what I'm doing with this t-shirt
1:27:33 design, just start changing the the look
1:27:36 and feel.
1:27:38 Okay, that doesn't look like the matrix
1:27:40 at all.
1:27:45 Um, it
1:27:48 needs to be
1:27:50 much freakier and
1:27:56 matrixier
1:27:58 matrixier. Uh, the
1:28:02 typography should look
1:28:06 menacing and
1:28:08 digital as if on a
1:28:12 screen. And the
1:28:17 background should be more visible
1:28:22 and there should be a boundary.
1:28:32 And the
1:28:35 image should be
1:28:38 transparent. All
1:28:41 right. Or even DPI count. Yeah,
1:28:45 exactly. Um, I have seen still images
1:28:48 turned up in motion video in other AI
1:28:51 apps. Can chat GPT do this? I've seen
1:28:54 still images turned into motion video in
1:28:56 other AI apps. Yes, it can. So, let's go
1:28:59 over. So, okay, within chat GPT itself,
1:29:03 it cannot. But in chat GPT, you'll
1:29:06 notice there's a an area here on the
1:29:10 left with, you know, chat GPT operator
1:29:14 and Sora. Depending on your
1:29:15 subscription, you may or may not see
1:29:16 these. I think for Sora, you you need to
1:29:19 be at least a $20 a month
1:29:23 subscriber. All right. And Sora is the
1:29:27 video
1:29:28 app. Yeah. Oh, so you see Sora. Okay.
1:29:31 So, so here's how you do that. You go
1:29:33 into Actually, let's go grab Let's go
1:29:36 grab We'll go grab one of these Bob
1:29:39 Sheets. All
1:29:41 right. We'll grab this one because it's
1:29:44 weird. And then we're going to go into
1:29:48 Sora. And Sora can make images, too. So,
1:29:51 I'm going to go into Sora and I'm going
1:29:53 to say I'm going to paste that in. I'm
1:29:55 going to
1:29:56 say
1:29:59 make an image of this man in
1:30:05 a
1:30:07 velour
1:30:12 tracksuit jogging
1:30:16 in Central
1:30:19 Park
1:30:23 with New Yorkers.
1:30:27 is snickering at
1:30:34 him. All right. And then we're going to
1:30:36 So we're going to do image. We're going
1:30:38 to do 3x
1:30:40 two and we're going to go
1:30:43 remix added to Q my
1:30:46 media. And so this should make our dude
1:30:51 Bob in a tracksuit in Central Park.
1:30:56 And then we'll turn that image into a
1:31:01 video as soon as it's there. Let's see
1:31:03 if we've got I think we got another
1:31:05 shitty t-shirt design here. Oh my god,
1:31:07 that's useless. That's [ __ ]
1:31:15 horrible. Make the background more
1:31:17 vibrant and transparent. I know. I knew
1:31:20 that was a bad prompt. All right. So,
1:31:21 let's just say uh let's go back
1:31:26 to a 2000s
1:31:32 design based or let's see
1:31:36 inspired by the
1:31:39 '9s design
1:31:45 layout. And we'll go back to Sora and
1:31:48 see if anything good is happening there.
1:31:50 We got nothing. Oh, 49%. All right,
1:31:53 Bob's coming. Bob's on his
1:31:55 way. Hey, Bob. Bob.
1:31:58 Bob. Yeah, Kyle, tell them what they've
1:32:03 won. They They haven't won anything,
1:32:05 Kyle. It's It's not a game show. Hey,
1:32:08 Bob, didn't you used to be on a game
1:32:10 show? Shut up. Shut up, Kyle.
1:32:15 [Music]
1:32:20 Oh, that's awesome. The people are
1:32:23 laughing at
1:32:27 him. All right, so let's let's take this
1:32:32 and we're going to say create video out
1:32:34 of it. That is awesome. We're going to
1:32:38 do 16
1:32:41 by9. That's [ __ ] awesome.
1:32:44 And then we'll say
1:32:46 uh
1:32:49 man jogs
1:32:52 forward
1:32:54 as
1:32:57 onlookers snicker at
1:33:00 him. He is
1:33:04 oblivious. He was a big
1:33:08 deal once.
1:33:17 By the way, one of the uh one of the
1:33:20 secrets to AI readiness that we talk
1:33:23 about in the AI salon is is the cycle of
1:33:25 AI readiness. And the first step in the
1:33:27 cycle is play
1:33:30 first. Why a velour jumpsuit? Why are
1:33:33 people snickering at him? Why did he
1:33:36 used to be a big deal? Just to [ __ ]
1:33:38 entertain myself. If you take this AI
1:33:40 [ __ ] too seriously, it's going to really
1:33:42 depress you. So, you might as well have
1:33:44 some fun with it. Here's the one thing I
1:33:46 know about it. No matter what you think
1:33:48 about it, and for anyone in here who's a
1:33:50 doomer and a [ __ ] hater and someone
1:33:52 who [ __ ] hates me because I love
1:33:56 AI, here's the only thing I
1:33:59 know. The only thing I know is that it
1:34:03 isn't going away.
1:34:06 And if it isn't going away, there are
1:34:10 exactly two
1:34:14 choices. I suppose there's a third one,
1:34:16 but it's not really a choice. You could
1:34:18 be completely ignorant of it. So
1:34:20 assuming you're not completely ignorant
1:34:21 that it exists, there are two
1:34:25 choices. You deal with it or you
1:34:30 don't. That's it.
1:34:34 If you don't deal with AI, it's going to
1:34:36 happen to
1:34:39 you. If you deal with AI, if you get
1:34:42 curious, if you start playing, if you
1:34:44 start building, if you start building
1:34:46 community with other AI friendly people,
1:34:49 AI optimists, AI creators, what you're
1:34:53 going to discover is that there is such
1:34:55 a massive [ __ ] opportunity right now
1:34:58 for like the next three
1:35:00 years. We're so early. As as much as it
1:35:03 feels like you're really behind, you're
1:35:06 not. So many people are actively
1:35:11 actively resisting dealing with this.
1:35:14 It's not going away. Two choices. Deal
1:35:17 with it or don't. And if you don't,
1:35:20 you're going to be the [ __ ] victim of
1:35:21 it. And it's going to
1:35:24 suck. You can still have opinions about
1:35:26 it. You can still [ __ ] hate the way
1:35:28 they trained it. You can still hate the
1:35:30 fact that they don't know what the [ __ ]
1:35:32 they're doing and what the implications
1:35:33 of all this stuff are. Like all that's
1:35:35 You can hate the fact that it
1:35:37 hallucinates and it can't do [ __ ]
1:35:40 good. But like what doesn't make any
1:35:43 sense to me at all is to ignore it. Just
1:35:46 deal with it. [ __ ] deal with it.
1:35:48 Learn it. Learn what it can do for you.
1:35:51 And what you're going to end up finding
1:35:53 is, "Oh, holy [ __ ] That was that's kind
1:35:55 of cool. I didn't I didn't realize I was
1:35:58 so talented.
1:36:02 They only hate AI because they've never
1:36:04 tried AI. Exactly. Joe goes digital. I'm
1:36:07 fin. I'm glad I finally gave in. GPT is
1:36:09 really helping me get through life.
1:36:11 Yeah. It's not just about making your
1:36:14 job more efficient. It's It can deal
1:36:17 with hard [ __ ] that you're managing
1:36:20 through. It can be a a therapist, right?
1:36:22 It can be a a friend, a companion.
1:36:27 How's your shirt from the 2000s? Let's
1:36:29 go look at that. It's got to be so
1:36:33 good. What's
1:36:42 the I can't see the
1:36:47 letters.
1:36:49 Make Okay. I want you to start over.
1:36:55 Write a creative
1:36:58 brief for a really
1:37:03 funny way to
1:37:06 lay this
1:37:09 out and maybe even an icon or two. Wait,
1:37:15 a uh image or two. Illustration.
1:37:20 Illustration or two.
1:37:25 All right, let's see what what it comes
1:37:27 up with. And then let's go back to Sora.
1:37:30 Sora, what's going on? My media. Do we
1:37:33 have a jogger yet? We
1:37:46 do. It's so bad.
1:37:55 It's the Joe Biden
1:38:00 pivot. Oh, they're so bad. Oh,
1:38:04 Sora. Oh my god, it's so bad. But you
1:38:08 can do it in Sora. What I would
1:38:10 recommend is Luma Labs. Lumalabs is
1:38:12 pretty good. Or Hedra is if you want a
1:38:15 character to talk, Hedra is really good.
1:38:21 hed.ai, but let's go grab let's go back
1:38:24 to my media and grab an image
1:38:27 here. All right, we'll do we'll do this
1:38:30 one because he looks like he's aware of
1:38:31 them. So, we're going to copy this
1:38:33 image. We're going to go to Luma
1:38:36 Labs. Go back to this thing. Um this
1:38:40 thing. Oh, it's still waiting your turn.
1:38:43 It looks like Lumalabs is hosed. Oh, you
1:38:45 know what? You know what? I bet I need
1:38:47 to reload the
1:38:49 page. YouTube comment from Chef
1:38:54 Kelly. My virgin ears. So much cursing.
1:38:57 Ah, you know what are you going to
1:39:00 do? Okay,
1:39:04 here's Okay, he opened it. So, you could
1:39:07 actually This is a usable video if you
1:39:09 reverse it, right? You could start
1:39:11 there. Have it Have it end where it
1:39:14 starts, right? Have it have it end.
1:39:23 Oops. Have it end
1:39:26 there. Can you reverse
1:39:30 these? Download picture and picture.
1:39:33 Modify. No. Never. All right. Let's do a
1:39:37 new
1:39:38 one. So, we're going to paste in in
1:39:42 Bob man.
1:39:48 Jogs as
1:39:51 jogs as
1:39:56 onlookers. Snicker. And then one of the
1:39:59 cool things about Lumal Labs, you click
1:40:00 on this little camera icon and it's got
1:40:04 it's got these um different camera shots
1:40:08 that you can just pop in there. So,
1:40:10 let's do let's do handheld. That'll make
1:40:12 it look more realistic.
1:40:15 And then the other things Luma Lab does,
1:40:17 whoever was asking, I forget who it was
1:40:19 that was asking about it. Now we know
1:40:21 why they're laughing.
1:40:24 Um, once the video's
1:40:27 created, you can make sound effects. So,
1:40:29 let's go make some sound effects here.
1:40:31 So, I'm going to go to audio. I'm going
1:40:33 to say
1:40:35 um
1:40:40 rustling of
1:40:43 fabric being
1:40:47 buttoned. I don't know if those are the
1:40:50 right words. We'll find
1:40:53 out. Oh, and you might not hear this on
1:40:58 YouTube. That's pretty good.
1:41:02 Yeah, that I mean that could totally
1:41:04 work. So nice. All right, creative
1:41:08 brief. U all right, let's
1:41:13 see. Um, sketch out the one you think is
1:41:20 best and make No, let's
1:41:24 see.
1:41:26 Create the one you think is best and
1:41:29 make the
1:41:31 background
1:41:33 transparent. Tik Tok pen. I'm going to
1:41:36 go. You got me wanting to try this.
1:41:39 Seriously, Joe, go just go just go make
1:41:42 an outline, make some scenes, make some
1:41:45 images, and then go dump them into
1:41:48 Lumalabs, Sora, just any anything. just
1:41:52 go find there there are like a dozen
1:41:54 different video tools right now and then
1:41:57 yeah you should share it in the salon if
1:41:58 you haven't joined the salon go to the
1:42:00 salon.ai AI, click on join our
1:42:02 community, and then in there there's a
1:42:04 showand tell area, but there's also an
1:42:06 AI learning lab irregulars channel down
1:42:09 in the clubs. Um, and and share share
1:42:12 what you create. All right. Oh, V2.
1:42:14 Yeah. Yeah.
1:42:17 Yeah. Oh, that's kind of nice. I like
1:42:20 this. What does this button do?
1:42:31 It's like being an idiot with
1:42:33 superpowers. It's like an idiot with
1:42:36 superpowers. Oh, they kind of got the
1:42:39 the thing wrong. Oh, and this is one of
1:42:41 the things I've noticed about um about
1:42:44 chat GPT
1:42:46 um transparent images is if it makes
1:42:50 something white like a talk bubble that
1:42:52 becomes transparent and you no matter
1:42:53 how much you describe it, it can't it
1:42:56 can't seem to [ __ ] understand that.
1:42:58 I'm going to say try again and make
1:43:05 any thing that is supposed to
1:43:10 be
1:43:12 white
1:43:15 slightly off
1:43:16 white so you
1:43:19 don't make it part of the background.
1:43:26 Also, your
1:43:31 spelling is a
1:43:33 [Laughter]
1:43:36 mess. All right, cool. All right,
1:43:38 thanks, Joe. Thanks. Seriously, thanks
1:43:40 for hanging out and thanks for the great
1:43:42 questions. Yeah, I'm going to get out of
1:43:43 here soon, too. Um, all right, cool. Cap
1:43:46 cut has reversed. Okay, good. Wait,
1:43:49 where was I running
1:43:50 to? Kyle is the Gordon Ramsay of AI.
1:43:54 [ __ ] right. I'm the I'm the Gary Vee
1:43:57 and Gordon Ramsay all all rolled up into
1:44:01 one. Oh man. Luma generate. Yeah. Let's
1:44:05 go see what we got. What we
1:44:08 got? Here we go.
1:44:14 [Laughter]
1:44:18 Okay, we'll do we'll do um uh
1:44:23 crowd heckles
1:44:26 and
1:44:29 heckles
1:44:31 heckles and
1:44:34 laughs
1:44:36 at jogger in central park.
1:44:42 This should be Sometimes these audios
1:44:45 are really bad, but this would make it
1:44:46 actually kind of funny. What's the site
1:44:49 for sound effects? Well,
1:45:01 [Laughter]
1:45:04 there's I know you can't hear it. Hang
1:45:06 on. Hang on. I'll I'll change the the
1:45:09 sharing because it's pretty funny. It's
1:45:11 pretty uh
1:45:13 nonsensical. Let's see. Ideas. Dream
1:45:16 Machine. Um Dream Machine. So,
1:45:20 Lumalabs.ai is where I made this image
1:45:23 to video and it's also where I added the
1:45:25 sound effects. You can also make
1:45:27 standalone sound effects in 11
1:45:30 Labs. But let's look at Bob. Poor Bob,
1:45:33 man. He was having a bad day.
1:45:37 [Laughter]
1:45:46 [ __ ]
1:45:47 [Laughter]
1:45:52 Bob. Oh my
1:45:56 god. Oh, never a dull moment. Never.
1:46:01 Never. What does this button do? It's
1:46:03 like being an idiot with superpowers. Oh
1:46:05 man, it's so close. Okay. Wait, what
1:46:08 just
1:46:10 happened? Oh, there we go.
1:46:13 Um, try one
1:46:16 [Music]
1:46:17 more time. Tabs, tabs, tabs. Oh,
1:46:25 yeah. I'm a little teapot, short and
1:46:28 stout. Here is my handle. Here is my
1:46:32 spout.
1:46:39 This one's pretty good, actually. What
1:46:41 does this button do? It's like being an
1:46:44 idiot with
1:46:45 [Laughter]
1:46:48 superpowers. Yeah, they added sound. I I
1:46:50 know. Like Lumalabs Lumalabs is adding a
1:46:54 lot of
1:46:55 stuff. Um, Hedra for charact for voice
1:47:01 animation. like if you record or
1:47:03 generate here's my new text stack for
1:47:06 characters
1:47:07 talking is I'll do mostly midjourney for
1:47:12 character development for the image
1:47:14 development then I go into hedra and I'm
1:47:17 using cartisia.ai AI for the voice
1:47:20 cloning. It's really good at voice
1:47:23 cloning. And then when you do the speech
1:47:25 to text, it's really good. The one thing
1:47:27 you have to do is you have to put
1:47:29 ellipses where you want pauses because
1:47:31 otherwise it talk. It runs everything
1:47:33 together. But um it's pretty [ __ ]
1:47:36 good. Pretty [ __ ] good. Bob didn't
1:47:39 win anything. No, Bob. Bob is being
1:47:41 heckled in Central Park and I don't
1:47:43 think it's appropriate. He didn't do
1:47:45 anything. Sure. Velour went out in the
1:47:48 80s. That's not it. He He's right. It
1:47:52 will come back around. It will come back
1:47:57 around. Pizza later. Hedra is wildly
1:48:00 good. Yeah, it it's shockingly good.
1:48:02 like I've been selling within my company
1:48:05 these these video personas, you know, to
1:48:07 to clients and things like that and and
1:48:10 the quality jump from what we had
1:48:12 before, whether it was Hey Jen or just
1:48:14 any of the tools, um, runway with with
1:48:17 the the acting thing in order to get
1:48:20 something that looked even remotely
1:48:22 decent, it it was just like, eh, it's
1:48:25 okay. And then Hedra or Hedra all of a
1:48:28 sudden is just like, it's so good. So
1:48:31 good.
1:48:33 All
1:48:34 [Music]
1:48:37 right, let's see if we got one. It's
1:48:40 like being an idiot. Wipe wipe
1:48:47 powers.
1:48:49 Let's
1:48:51 simplify
1:48:52 the
1:48:55 punchline to
1:48:59 idiot. What do
1:49:02 I'm an
1:49:06 idiot
1:49:08 with super
1:49:11 powers lowerase A,
1:49:19 right? Inpaint it. I haven't had very
1:49:22 good luck with inpainting at all in chat
1:49:24 GPT. I find that useless at this point.
1:49:28 Damn, that's not Central Park. That's
1:49:30 the west
1:49:34 side. That's pretty
1:49:36 [Laughter]
1:49:39 funny. Oh, it is. You're right. It's on
1:49:42 the Hudson River. Look at
1:49:45 this. Yeah. See that? That's Jersey over
1:49:48 there. That's Jersey. That's the Hudson
1:49:51 River, [ __ ]
1:49:52 [Laughter]
1:49:59 Yeah, I agree with that. It doesn't seem
1:50:01 to support in painting. It always makes
1:50:02 a new image. I bet they add it soon. I
1:50:05 hope they add it soon because it's like
1:50:06 the the image gen is so good, but it's
1:50:09 like get it together. The other thing
1:50:11 that I know that they're aware of is the
1:50:14 image gen tends to have this yellow cast
1:50:17 to it, and I think they're fixing that
1:50:19 as
1:50:21 well. Um, all right, one more. AI. What
1:50:24 does this button do? I'm an idiot with
1:50:26 superpowers. Oh my god. S U P O R. Well,
1:50:31 there you go. And listen, I know those
1:50:34 of you with Photoshop skills. I've got
1:50:36 them, too. Are like, couldn't you just
1:50:38 do that in Illustrator? Make the letters
1:50:39 in Illust? Yeah, you could. You could.
1:50:43 But now I'm lazy. I don't want to use my
1:50:45 brain. I don't want to use my
1:50:47 skills. I want to use my
1:50:49 magical
1:50:51 superpowers like the
1:50:53 [Laughter]
1:50:57 t-shirt. You've had luck with in
1:50:59 painting if I say so. If you say what?
1:51:04 Change. Nothing else. All right, we'll
1:51:07 try
1:51:09 it. Producer Brandon claims claims this
1:51:12 is going to work.
1:51:20 Um, fix the
1:51:24 spelling
1:51:27 from
1:51:33 supers
1:51:35 to super
1:51:40 powers change nothing else.
1:51:45 in the
1:51:47 image. All right, we're going to try it.
1:51:51 Bets. Who wants to bet there? Wagering
1:51:54 is allowed on the on the on the AI
1:51:56 learning
1:51:59 lab, but not on Tik
1:52:01 Tok. Oh, yeah. Actually, you can't say
1:52:04 that. I'll probably get banned for life.
1:52:06 In my experience, it'll get close with
1:52:09 the composition in the reproduction, but
1:52:11 it's still a new image. I've got five on
1:52:14 it. On what it doing it? I'll take that
1:52:16 bet all day long. I bet five roses. All
1:52:18 right. Yeah. Bet the number of roses.
1:52:21 And And if it [ __ ] it up, you have to
1:52:24 pay that amount of
1:52:25 [Music]
1:52:27 roses. It's not supported to my
1:52:29 knowledge. Yeah, I don't think it is. I
1:52:32 don't think it is. But what's weird is
1:52:35 like it's gotten like Well, I don't know
1:52:38 if that's the case.
1:52:40 the the image generation stuff in
1:52:44 editing has always kind of sucked. So if
1:52:46 if they actually haven't it's possible
1:52:48 that this is this is the old image model
1:52:50 is that's probably what's going
1:52:52 on. Although look it it is making it
1:52:56 with the background. So let's see let's
1:52:58 see what it does
1:53:00 people. If producer Brandon's right you
1:53:02 know I'm never going to hear the end of
1:53:04 it. You know you Hey, you remember that
1:53:06 time back in April 2025 when I was super
1:53:09 right and you were like super wrong
1:53:10 about something? Remember
1:53:14 that? Uh oh. Oh no. Did he get it right?
1:53:18 Please, please, please do not
1:53:25 Ah, I thought I thought I was on the
1:53:28 hook there. It it it did change it a
1:53:30 little bit and it did mostly make the
1:53:33 thing the the same, but it's like it
1:53:35 changed the colors. So anyway, all
1:53:38 right.
1:53:41 Whatever. I I really do like this
1:53:43 t-shirt. Let's see. Try one more time.
1:53:45 Try one more time. Your
1:53:51 spelling sucks.
1:53:56 I'm an
1:53:58 idiot
1:54:01 with
1:54:05 super
1:54:12 powers. I'm going to leave all the extra
1:54:15 L's in my spelling my in my insult.
1:54:23 [Laughter]
1:54:28 All right, people. I'm getting out of
1:54:29 here. All right, let's We're gonna This
1:54:31 is going to be the last thing. This
1:54:32 gonna be the last thing. I'm going to
1:54:35 go. We're going to be We're We're ger
1:54:38 Gordon Ramsaying on out of here. All
1:54:42 [Laughter]
1:54:45 right. Oh, man. Yeah, tomorrow is just a
1:54:49 normal Tuesday.
1:54:52 There's no
1:54:53 salon. Oh, and next Tuesday I am on a
1:54:59 plane. I'm going to need to find a
1:55:01 co-host for next
1:55:04 Tuesday. Wait, is it next Tuesday? Oh,
1:55:06 it's not. No, it's the following
1:55:07 Tuesday. No, I'm fine. Next Tuesday is
1:55:09 not a salon
1:55:10 either. But I'm I've been So next week,
1:55:14 is it next week? Yeah, next week from
1:55:17 Tuesday to Thursday I'm in Washington DC
1:55:20 doing like AI
1:55:22 [ __ ]
1:55:24 Um, the AI salon is sponsoring a big
1:55:27 thing. Did it get it right? Nope, it
1:55:30 didn't. I'm an idiot with
1:55:33 superpowers. I
1:55:42 wonder. No,
1:55:44 whatever. We're done. We're out of here.
1:55:48 All
1:55:50 right. Um, so what's coming up this
1:55:54 week? I've got a podcast Wednesday. Uh,
1:55:56 400 PM Mountain time is the AI readiness
1:55:59 project with myself and Anne Murphy. Um,
1:56:01 if you're on YouTube, or even if you're
1:56:03 not, please like and subscribe on
1:56:05 YouTube. Tik
1:56:07 Tok could go away any day. Like, we
1:56:10 don't know. Like, there's so much
1:56:11 [ __ ] chaos going on right now. We
1:56:13 should just assume that at some point we
1:56:15 won't be able to live stream on Tik Tok.
1:56:17 So,
1:56:19 um,
1:56:20 subscribe on the YouTube channel. Uh,
1:56:23 it's learning
1:56:25 lab-ai on
1:56:27 YouTube. All right.
1:56:31 Beautiful, beautiful, beautiful,
1:56:33 beautiful. All right, peace out
1:56:35 everybody. Peace out.
1:56:39 [Music]