AI Learning Lab

3/31/2025 - AI Adventures: Exploring Runway Gen 4 and Zeni Eyewear

FrJv8C-C_cc
Live Stream2025-04-011:40:1777 views

Description

This AI Learning Lab session dove into the exciting world of AI video and image generation, focusing on RunwayML's new Gen-4. The host, Kyle Shannon, explored its touted "world consistency" feature, aiming to create a video with a consistent character across scenes. However, technical difficulties and UI confusion led to a humorous and frustrating deep dive into the platform's complexities, highlighting the challenges of using cutting-edge AI tools. The session also touched upon generating print ads with AI, using Chat GPT to create mockups inspired by existing designs. Beyond RunwayML, the conversation explored other AI tools like Sora, Higsfield.ai, and Chat GPT's image generation capabilities. Kyle experimented with creating a magazine ad, generating video from still images, and even transforming a photograph into a Lego kit concept. The discussion also briefly covered AI voice generation tools like 11labs and Cartisia.ai. Learn more about AI on TikTok: https://tiktok.com/@aiLearningLab. #AI #ArtificialIntelligence #VideoGeneration #ImageGeneration #RunwayML #ChatGPT #AICreativity #AIArt Chapters: 00:00:00 Introduction 00:00:43 Addressing the Dog 00:01:00 Song Interlude 00:02:00 Wardrobe Malfunction 00:03:12 Producer Brandon's Return and New Glasses 00:04:25 Ramy's Attitude 00:05:38 Call to Action on TikTok 00:06:00 Questions and Answers 00:06:37 Another Song Interlude 00:07:41 OpenAI's Open-Source Model 00:08:49 Zeni Glasses Sponsorship 00:09:34 Creating a Magazine Print Ad 00:10:40 Technical Difficulties 00:12:02 Zodiac Jokes and Ad Creation 00:15:50 Ad Copy and Concept Generation 00:17:30 Refining the Ad Concept 00:18:46 Ad Review and Feedback 00:20:15 Runway Gen 4 and Video Generation 00:21:40 Character Creation with ChatGPT 00:24:31 Elliot on the Edge Video Generation 00:25:34 Runway Gen 4 Confusion 00:28:02 Runway Character Consistency Issues 00:32:09 WebP Format Rant 00:34:20 Further Runway Experimentation 00:36:39 Magazine Ad Critique 00:37:18 Another Runway Attempt and Frustration 00:39:41 GIF Pronunciation Debate 00:41:43 Carpet Salesman Kyle 00:43:05 Zeni Ad Review 00:45:04 Explaining Runway Gen 4 Features 00:49:04 More Runway Struggles 00:52:00 Reading the Manual 00:53:50 More Interface Issues 00:55:08 Drag and Drop Failure 00:58:18 Giving Up on Runway 00:58:48 AI Salon Promotion 01:00:00 Upcoming AI Salon Presents Guests 01:01:04 Sora and Image Generation 01:02:39 Sam Altman and Image Gen Feature 01:03:09 AI Literacy and Blot Presentation 01:04:55 Voice AI Recommendations 01:06:00 Domain Name Rant 01:07:59 Toilet Paper Patent 01:09:36 Higsfield.ai Exploration 01:13:26 Higsfield.ai Image to Video Test 01:16:45 Video Tools Discussion 01:18:00 Image Generation and Website Mockup 01:22:00 April Fool's Day and Closing Remarks 01:24:41 Lego Kit Creation with ChatGPT 01:28:21 Toilet Paper Orientation Discussion 01:31:00 Final Thoughts and Sign-Off

Chapters

0:00Introduction0:43Addressing the Dog1:00Song Interlude2:00Wardrobe Malfunction3:12Producer Brandon's Return and New Glasses4:25Ramy's Attitude5:38Call to Action on TikTok6:00Questions and Answers6:37Another Song Interlude7:41OpenAI's Open-Source Model8:49Zeni Glasses Sponsorship9:34Creating a Magazine Print Ad10:40Technical Difficulties12:02Zodiac Jokes and Ad Creation15:50Ad Copy and Concept Generation17:30Refining the Ad Concept18:46Ad Review and Feedback20:15Runway Gen 4 and Video Generation21:40Character Creation with ChatGPT24:31Elliot on the Edge Video Generation25:34Runway Gen 4 Confusion28:02Runway Character Consistency Issues32:09WebP Format Rant34:20Further Runway Experimentation36:39Magazine Ad Critique37:18Another Runway Attempt and Frustration39:41GIF Pronunciation Debate41:43Carpet Salesman Kyle43:05Zeni Ad Review45:04Explaining Runway Gen 4 Features49:04More Runway Struggles52:00Reading the Manual53:50More Interface Issues55:08Drag and Drop Failure58:18Giving Up on Runway58:48AI Salon Promotion1:00:00Upcoming AI Salon Presents Guests1:01:04Sora and Image Generation1:02:39Sam Altman and Image Gen Feature1:03:09AI Literacy and Blot Presentation1:04:55Voice AI Recommendations1:06:00Domain Name Rant1:07:59Toilet Paper Patent1:09:36Higsfield.ai Exploration1:13:26Higsfield.ai Image to Video Test1:16:45Video Tools Discussion1:18:00Image Generation and Website Mockup1:22:00April Fool's Day and Closing Remarks1:24:41Lego Kit Creation with ChatGPT1:28:21Toilet Paper Orientation Discussion1:31:00Final Thoughts and Sign-Off

Transcript

0:00 [Music]
0:14 You're very loud. You're a very loud
0:17 doggy. You're a very loud
0:20 doggy. Very loud. You know you're loudy,
0:22 right? You know that. Do you know you're
0:25 loudy? You know you're loudy.
0:29 [Music]
0:35 [Music]
0:43 Hey, hey, hey, hey. No rapping. No
0:47 rapping.
0:49 [Music]
1:00 easiest place I can raise my whole
1:04 [Music]
1:07 head my thoughts in sweet
1:11 [Music]
1:14 silence and this is place where the
1:17 feelings are
1:19 [Music]
1:21 dead from an and exposed to
1:25 violence. This is place I can slowly
1:29 face. The only one I truly can
1:34 know. These are tears from a long time
1:38 ago. Got these tears from a long time
1:42 ago. I need to cry 30 years or
1:46 so. These are tears from a long time.
1:54 [Music]
2:00 Go. Oh darling, oh darling, you say unto
2:07 me, where have you been all my
2:12 liime? Oh, that's a bad shirt for Tik
2:14 Tok. Let me flip the camera. That's
2:17 better. And I have been swimming the
2:20 seven sad
2:24 seas. Four women have cost me their
2:28 lifelines. Well, I to the bottom
2:33 ster I didn't know.
2:37 These are tears from a long time
2:40 ago. Got these tears from a long time
2:44 ago. I need to cry. 30 years old.
2:49 So these are tears from a long
2:54 time
2:56 [Music]
3:13 ago. Good evening good people. What is
3:15 happening? Good day to you
3:20 all. What is going
3:23 down? What is happening? What's he
3:26 happening? What's happening? Producer
3:27 Brandon is back indeed. We got producer
3:30 Brandon back. I've got new glasses,
3:32 which is in theory a good thing, but
3:35 I've been wearing the wrong glasses for
3:37 long enough that my eyes are all
3:39 maladjusted. So, if I seem like I'm like
3:42 [Music]
4:02 new glasses. Congrats. Thank you. It's a
4:05 It's a big It's a big moment in my life.
4:12 [Music]
4:25 You're Ramy tonight. You're Rammy
4:27 tonight. I don't know what's going on.
4:28 You got a lot of attitude. Got a lot of
4:31 attitude for a fat dog. It's a lot of
4:34 attitude.
4:37 [Music]
4:40 look like your grandma's glasses. Thank
4:42 you, Mr. It. That's terrific.
4:48 [Music]
5:09 I can't I can't I just can't even play
5:11 anymore. I can't even play anymore.
5:13 You're too loud. You're too loudy
5:14 tonight. I don't know what your deal is.
5:17 What is your deal? Huh?
5:21 [Music]
5:39 Do me a favor. Do some some liking on
5:41 Tik Tok and maybe share it if you can.
5:44 We've got 12 people in there. Like Tik
5:48 Tok, please.
5:56 [Music]
5:59 Do I ever play in Denver? I do not. I
6:02 don't play outside of this office.
6:06 [Music]
6:15 [Music]
6:37 Standing
6:39 between you and a hard place is insane.
6:46 Standing too
6:48 near you in a fire makes it
6:55 clear. Your trouble to
7:01 me. Real trouble. Can't you
7:06 say leading in
7:09 close? Smell of your perfume scares me
7:12 most.
7:16 Leaning
7:18 away. It feels stronger every
7:24 day. Your trouble to
7:30 me. Real trouble. Can't you see?
7:41 That's Champ Shannon on the uh closing
7:43 out the
7:45 [Laughter]
7:48 song. So, good day good people.
7:51 Um couple of things from today worth
7:54 talking
7:56 about is
8:01 um OpenAI announced today they are going
8:04 to release an open-source model with
8:07 weights
8:08 um sometime in the coming months and
8:11 they want to know from
8:13 developers what they want in
8:16 that and I'm assuming developers are
8:20 going to
8:20 put all the [ __ ] in
8:25 it. But, you know, it it it should
8:28 probably at least be as good as
8:30 DeepSeek. You know, the the the Chinese
8:33 one that that was better than
8:36 GPT40 and they open sourced it and it
8:39 caused $1.2 trillion dollars in stock
8:41 market bleeding. That one should
8:44 probably be that good.
8:50 Check out the new specs. I know, right?
8:52 Zeni, baby. I should This should be a
8:55 sponsorship deal. Hi. You know, I was
8:58 squinting for years. Then I went to Zeni
9:01 and I got these fine
9:03 spectacles. Mr. It says they look like
9:05 his
9:06 [Laughter]
9:09 grandmother's. It hurts.
9:14 Zeni, for your all your I care
9:19 needs. We could make a commercial
9:22 tonight. We could go make a print ad,
9:24 couldn't we? Watch. Let's go do it.
9:26 Let's go make a print
9:27 [Laughter]
9:35 ad. Uh, let's see. Uh, make let's see. I
9:41 want to
9:44 create a
9:46 powerful
9:49 magazine print
9:51 ad for the
9:55 likes of
9:58 Vogue. Oh, you know what we'll even
10:01 do? Wait, how am I going to do this? Oh,
10:03 I know how I can do
10:05 this. Watch this. This is going to be
10:07 fantastic. I know. Mirror infinity
10:09 mirror. Shut up. It's gonna be okay. All
10:11 right. Uh, we're gonna do
10:23 screenshot. Shang dang dang dang dang
10:26 dang. I don't know what anyone can see
10:29 right now, but that doesn't
10:32 matter. Just calm down. Just everybody
10:36 calm down.
10:38 Everybody's getting
10:40 Ramy. Dog was Ramy. You're all Ramy. We
10:44 can't We can't see anything. Kyle. Hey,
10:46 Kyle. Kyle, I don't know what you're
10:48 doing. If I can't see your screen, I
10:50 don't know what you're doing. Kyle.
10:52 Kyle. Hey. Hey, Kyle. Kyle. Kyle. Kyle.
10:56 Kyle. Calm down. Just calm
11:04 down. We see.
11:07 We see
11:09 scenery. I see dead people. Am I going
11:12 to do this to clipboard? No, I'll do
11:13 this to desktop. Okay. Looking at the
11:16 camera. Okay. So, so here's what I did.
11:20 It's going to be You're going to be
11:21 like, "Oh my god, that was so
11:23 creative. That's so
11:26 creative. That was so creative." He's
11:29 creative. He's pretty
11:31 creative. Uh oh. What did I
11:34 do? Dang it.
11:43 I screwed up the white the black box
11:46 thing cuz these stupid blackbox sites
11:49 are
11:54 stupid. No, that's not quite black now.
11:56 It's kind of blue. Hey, Kyle. Hey, Mimi.
11:59 What's
12:00 happening?
12:02 Okay, Ramy. Well, it is the Aries
12:05 season. Oh, we're doing zodiac jokes
12:07 tonight. I like it. Okay, let's share
12:10 let's share our screen
12:12 again. We're gonna do some action. We're
12:15 gonna
12:20 [Music]
12:24 do Stop freaking out. Listen to Josh.
12:28 Evening, guys. Archetypal architect
12:30 showed up. It's okay, everyone. He's
12:32 here. We're good to go. All right. I
12:35 want to create a powerful magazine print
12:36 ad for the likes of Vogue,
12:40 GQ, and
12:45 um Agriculture.
12:52 Agriculture uh list
12:55 [Laughter]
13:01 monthly, proving once again that the AI
13:04 learning lab is really just an
13:05 opportunity for Kyle to entertain
13:07 himself. Thank you. Um the ad
13:12 will be for
13:16 Zeni Eyewear. I
13:20 wear uh
13:22 site. Okay, wait. We got to go find the
13:25 Zeni site,
13:27 right? Uhuh. Uhuh. Zeni. And then we got
13:30 to get the logo from Zeni. Zeni Optical.
13:34 All right.
13:36 Fine.
13:39 Copy. Back to here. Put
13:43 zenioptical.com. And then let's go get
13:45 the logo, shall
13:48 we? Uh, wait, why? Oh, I've got to
13:52 accept
13:53 cookies. Thanks, marketers. Save image
13:57 as Zeni
13:59 logo desktop. So, we're going to upload
14:02 Zeni logo. Oh, no, we don't need that
14:05 yet. We'll we'll get to that in a
14:07 second. Um,
14:10 uh, and let's see. And let's see.
14:16 Um, I
14:19 want the
14:22 target
14:28 audience of the
14:33 ad to
14:35 be what do we call them? um AI
14:41 optimists and the nickname for
14:46 them
14:48 is the irregulars.
14:57 They like to
15:01 say, "We're not
15:04 weird. We're
15:10 irregular." Which
15:13 everyone
15:15 knows is not
15:18 true. But them, they are
15:25 weird and
15:27 [Laughter]
15:33 lovable. That's
15:35 good.
15:37 Write me the
15:41 copy and concept for five ads for this
15:48 target
15:50 audience. All right. So, we're off.
15:53 We're making we're making we're not
15:55 using our brain. So, we used our brain a
15:58 little bit to write that prompt, but
15:59 that was really just stream of
16:01 consciousness, right?
16:03 See through the noise alone irregular
16:07 thick chic nerd. Wait. Oh, think chic.
16:12 Nerdchic like fashionforward AI
16:16 tinker stands in a glitching pixelated
16:19 crowd. I don't know what that mean.
16:20 Whatever. It doesn't matter. Again, we
16:23 don't need to use our brain here. We
16:25 just need to read these and if anything
16:26 sounds good, we'll make that. All right.
16:29 Fashionably untrained. I like this. A
16:32 runway model robot. No, we don't need
16:35 robots. Um, ad concept. We are not
16:38 weird, but we are. A split image side.
16:40 Chaotic whiteboard scribbles. Lab
16:42 goggles. A rooster in sunglasses.
16:45 Agricultural monthly.
16:50 Hello. The other side. High fashion
16:53 black and white portrait of the same
16:55 person in zeni glasses looking intensely
16:58 editorial. We're not weird. We're
17:00 irregular. Everyone says we're weird. We
17:03 say define
17:05 weird. That's pretty
17:07 good. Zeni frames the world through
17:11 lenses made for visionaries. All right.
17:13 Eyes on the
17:14 horizon.
17:17 Um, frames of mind. A surreal collage.
17:20 Shakespeare's bust. I I don't know. It
17:23 doesn't matter. Um, okay. I want you
17:30 to
17:35 pick the concept and
17:41 feature the
17:45 person here
17:48 wearing zeni
17:51 glasses and use their logo.
17:58 Go make sure the ad will win
18:04 awards. And then we're going to upload
18:06 the screenshot of me I just took in my
18:09 Zeni
18:11 glasses. There's the Zeni logo.
18:15 Open. And then we're going to go upload
18:18 from
18:20 computer. No. No.
18:25 Oh, different desktop.
18:28 Screenshot name contains
18:33 screenshot. That's not right. Where is
18:36 it? Where is
18:37 it? That's it. There it is. Look how
18:41 good I look. Look how
18:43 fantastic. Fantastic. Bob, tell him what
18:46 he's
18:47 won. All right. Boom. Let's Let's have
18:50 this thing make an ad for us. Excellent.
18:54 You've got a regular human irregular
18:56 front and center. Look, rocking zeny
18:58 frames with a charming glint of weird
19:01 but lovable energy based on the vibe in
19:03 the photo. Oh, this is
19:07 interesting. And the goal to make this
19:09 an award-winning print ad, I'm choosing
19:11 ad concept 3. We're not weird, but we
19:13 are. Because it's personal, cheeky, and
19:16 instantly connects to your
19:18 crew. Here's the full print ready ad
19:21 print concept. Let's see. Target Vogue,
19:24 GQ, and Agriculturalist
19:29 Monthly. Target audiences, the
19:32 irregulars, AI optimists who wear their
19:35 weird proudly. Ad type full page print
19:38 ad featuring the real radiant irregular
19:41 from the
19:43 photo. Zeni included below.
19:47 Okay, we're not weird. We're irregular.
19:52 Um, let me know if you'd like this tuned
19:56 turned into a full print layout mockup
19:59 with typography. Yes. Make the
20:10 ad. All right. So, that should be
20:13 fun. So, while we're at that, let's go
20:16 over to
20:17 runway and we're going to go to
20:21 New session, generate
20:23 video. And we've got Gen 4 here. Now,
20:26 Gen 4 is
20:31 apparently it's got drag an image or
20:34 click to
20:36 upload. So, we're going to upload. Why
20:39 don't we just upload that screenshot I
20:40 just took of
20:43 me? Oh, this is all about you. Is this
20:46 all about you?
20:47 I was actually thinking about why I
20:50 choose me all the
20:53 time. And then you're going to go, "Uh,
20:55 the answer's obvious, you big [ __ ]
20:57 egoomaniac. You're the one with the
20:59 [ __ ] channel." I that fair fair fair
21:03 fair fair fair fair fair fair fair fair
21:03 fair fair fair fair fair fair fair fair
21:03 fair fair fair fair fair fair fair fair
21:03 fair fair fair fair fair fair fair fair
21:03 fair fair fair fair fair fair fair fair
21:03 point but
21:09 um when you do a thing of you you know
21:13 if it looks like you like I did some
21:15 things today in chat GPT and they didn't
21:16 look like me and you kind of instantly
21:19 know if you do it of someone else you
21:21 kind of justify it they sort of look
21:22 like
21:25 that
21:31 Okay. Did you upload? Okay, we uploaded
21:34 it. All right. So, we're going to say,
21:36 All right. So, now we're going to go to
21:37 another chat jeep session. Let's go to
21:40 here and go new chat
21:43 jeep. We're gonna say,
21:45 um, I want you to come up with a
21:53 memorable character in a
21:56 [Music]
21:57 memorable
22:00 location in the middle of memorable
22:05 action.
22:09 It should
22:11 be
22:18 cinematic
22:19 and
22:21 describe a main
22:26 character who wears glasses. No, a main
22:30 character.
22:34 what
22:36 they are wearing. Can you see all this?
22:40 Yes.
22:43 Um
22:46 and what they are
22:50 doing.
22:53 Uhuh. The bell ringer of Black Hollow.
22:57 No, sorry. Stop.
23:00 It's a
23:02 middleaged man with
23:06 glasses. Work with
23:09 that. Make it more
23:16 modern. Absolutely. Let's sharpen
23:19 it. The man on the
23:23 billboard. The character Elliot Vain.
23:26 Get it? A middle-aged man in a rumpled
23:29 navy suit that's one size too big. Like
23:31 he borrowed it from a more confident
23:34 version. Setting on top of a 20story
23:37 billboard in Time Square. Blinking ads.
23:40 The crowd doesn't notice him. Not yet.
23:42 The city roars on.
23:52 Um, this just needs to be
23:58 a visual, no
24:01 dialogue or real
24:05 action. Um, it should center on the
24:12 person, not the location.
24:18 [Music]
24:20 Elliot on the
24:31 edge. All right, this is
24:34 good. So, here's a description. That
24:37 looks
24:39 fine. We're going to go back to this
24:42 place.
24:44 Runway Gen
24:45 4.
24:48 Oh, let's see. What did it cut
24:57 off? Okay, he doesn't move. He doesn't
24:59 need to. He is the scene. That's
25:04 fine. And we'll do 10 seconds. So, it
25:08 should make it look like me. So they
25:11 claim that there's good character
25:12 consistency
25:14 here. The the
25:19 agricultureonline.com. Thanks
25:21 [Laughter]
25:24 HT. Um three is going to be my choice as
25:26 well. No wonder my writing
25:29 AI sounds like other AI writing. That's
25:32 funny. U take
25:35 two. You guys are
25:37 good. Side hustle Mimi putting the kids
25:39 to bed. Very nice. Oh, let's go back to
25:41 chat GPT and see how that image came
25:48 out. Okay, but that's really bad because
25:51 you can't read the copy.
25:56 Um, okay. That doesn't look like an
26:02 award winning
26:06 ad. He looks frumpy.
26:10 which he is. And the
26:15 shirt sucks. And you can't read the
26:20 white text on the white
26:26 shirt. And
26:29 the title copy t the title
26:38 typography is
26:41 hack is hack.
26:46 If if you were my art
26:52 director, this would be your last
26:57 project. Try
26:58 [Laughter]
27:07 again. Looks kind of like trying to read
27:10 the Tik Tok comments in the normal Yeah,
27:13 in the normal live. Yeah, exactly.
27:17 Exactly. You should tell it to add a
27:20 black bar. Well, I want to see if it can
27:21 figure out it out on its
27:23 own. All
27:25 right. Here's So, wait. The description
27:29 here
27:33 is a man stands
27:35 alone at the edge of a rooftop parking
27:38 garage just after dusk. His silhouette
27:40 is sharply defined against the low
27:42 orange glow of the city. He's in his
27:44 late 40s, maybe early
27:46 50s.
27:48 Um, soft around the middle. That's
27:51 [ __ ] rude. With a once meticulous
27:54 haircut, grown out just enough to
27:56 suggest something is
28:03 unraveling. This is This is rough,
28:05 people. His charcoal gray suit jacket is
28:09 slung over his silhouette by a single
28:11 finger, wrinkled, forgotten. His light
28:14 blue dress shirt clings to him in the
28:16 humidity. Collar open, tie missing.
28:20 Thinframed glasses sit low in the bridge
28:22 of his nose, catching the last bit of
28:24 neon from the nearby
28:26 sign. His eyes aren't sad, they're
28:28 quiet, just like something distant, or
28:31 maybe just lost in their own reflection.
28:33 All right. So, me on a parking garage at
28:35 sunset. That's This should be a
28:37 10-second
28:38 video. That should be
28:45 awesome. Uh, that doesn't look
28:55 [Laughter]
28:58 right. Wait, I I'm confused.
29:06 Well, that's kind of that's actually
29:08 kind of
29:11 cool. Oh. Oh, damn it. Damn that
29:20 thing. So, there's so there's a little
29:23 bit of traffic moving here. He's not
29:25 Okay. I clearly don't know how to use
29:27 this.
29:30 Oh, and it's 1.1 aspect ratio. So, that
29:33 was a loser loser move on my
29:43 part. Okay. So, so what do we do here?
29:47 We go we must
29:49 go image
29:59 reference add new.
30:05 No
30:11 frames
30:16 later. There's the prompt references
30:22 style. What just happened?
30:30 aesthetic range settings
30:40 style. Well, you know, we thought you
30:42 knew how to use this
30:47 thing. I'm so
30:50 confused. Anybody know how to use Oh,
30:53 damn it.
30:57 Whatever Apple did with their with their
30:59 window management in this last OS
31:02 upgrade. Good
31:05 lord. Rubble rubber rooster media
31:07 watching from the gym. Well, hopefully
31:09 you can't see it because I can't I don't
31:11 know what I'm doing.
31:22 Um, aspect ratio 16
31:27 by9.
31:30 Fine. But that's not what I
31:37 wanted. And I can't use How are they
31:40 getting character consistency?
31:44 What are you trying to do? What I'm
31:46 trying to do is Gen 4 apparently has
31:49 character
31:51 consistency. Like you can take a
31:53 character and put it in lots of
31:54 different scenes and it maintains that
31:55 character across scenes, but I don't
31:57 know how to add the character to the
32:00 scene because I don't know what I'm
32:01 doing because I didn't do this before
32:04 coming
32:05 [Laughter]
32:09 here. Let's see. What's that say?
32:22 Dude, did you just pronounce us as a
32:24 word?
32:25 Yes. Us. Your operating system. Shut up.
32:29 Don't. What do you Next you're going to
32:31 ht. Next you're going to tell me you
32:32 pronounce it
32:34 GIF. If if we get into this argument
32:37 again, I mean, I'll go back to 1996.
32:40 I'll start this over.
32:42 I'll I'll do it. I'll go back
32:46 there. WEBP
32:48 format. We we we we came up with the
32:51 GIF, then we did the JPEG, then we did
32:54 the PNG. Everything was fine. JPEG
32:56 couldn't have transparency, so it needed
32:58 PNG. Fine. I understand the reason for
33:00 that. The [ __ ] is WEBP and then
33:03 AVIF? What What are we doing with our
33:06 lives? And why doesn't why do I why do I
33:09 not know how to use
33:12 this? I don't get
33:15 it. Fixed
33:20 seed. We don't need no stinking
33:23 manual. Did you read the manual? Of
33:26 course I didn't read the manual. It's
33:28 not how you do this. Everybody knows
33:31 that the most important thing to
33:35 do. It's aer them. I'm hearing a little
33:38 Frank. Shut
33:41 up.
33:43 Um
33:48 frames. How are you supposed to use a
33:50 reference
33:52 image? How? Reset all settings. Reset
33:56 all settings. Ah, type a
34:01 prompt or drop an image to turn into a
34:04 prompt.
34:21 Well, so there it's analyzing the image
34:23 of
34:29 me. Is it generating it?
34:45 So, you upload an image, it describes
34:52 it. I'm completely confused now.
34:58 I'm in the creative partner program for
35:00 Runway and I clearly haven't used it
35:07 enough. I got into three creative
35:09 partner programs all at the same time
35:12 and then I ended up just doing other
35:14 [ __ ] So that's bad. All right. Well,
35:17 but this is why we're doing it tonight.
35:19 Look at look at sort of me with sort of
35:22 shelves behind me.
35:30 All right. And so it says, oops, where
35:33 did that
35:34 go? So I can say, use this to make a
35:45 movie. View the
35:47 guide. Oh, there is a manual, Vicki. You
35:50 could read the manual. I don't like
35:53 that.
35:56 [Music]
35:57 Um, but I don't want that layout. I just
36:00 want to use a
36:03 dude. Wait,
36:05 create this first image you want in
36:10 chat. It will always start your video
36:13 with your reference frame. Maybe. Yeah,
36:14 that makes sense. All right.
36:19 So, good lord.
36:23 So, we got to go back to chat. JPEG
36:26 tag. Well, at least it put me in a
36:29 turtleneck
36:30 [Laughter]
36:39 now. Supposed to be a magazine ad.
36:42 What's with all the words? This We're
36:44 not reading a short story.
36:48 Um, cut. Uh, well, we're not gonna [ __ ]
36:51 that. We're just [ __ ] it. [ __ ] it. We're
36:54 starting over. We'll We'll do it live.
36:57 We'll do it
37:03 live. Uh oh. What did What did Vicki
37:06 find me? Did Vicki find me? I can't copy
37:09 it anywhere. Why can I not copy
37:12 that? Come on, Streamyard. Be suckless.
37:19 All right, I'm going to upload a picture
37:20 of
37:22 me and I'm going to
37:27 say man
37:30 stands on the
37:33 roof of a
37:36 parking
37:38 garage overlooking the
37:43 city at sunset.
37:47 at with
37:49 his
37:51 blue
37:54 blazer
37:55 slung over his
37:59 shoulder with
38:01 a
38:06 finger. Um, he's a little
38:11 soft. I like that term. He's a little
38:14 soft in the middle.
38:21 But once had
38:23 [Laughter]
38:28 something shot should
38:32 be 16
38:36 by9 and uh let's see
38:42 um cinematic
38:46 photography as if a still from a
38:51 movie. Hyper
38:55 realistic. Okay. I met the inventor of
38:58 GIF while pumping gas. His name was
39:00 Steven White developing CompuServe. But
39:03 wait, real estate in Columbus. If you
39:06 met the inventor, did you at least ask
39:07 him how is it? How is it pronounced? And
39:10 is it GIF or is it GIF? It better not be
39:13 GIF or I'll just I'll end the show
39:15 tonight. That'll be
39:17 it. If the inventor said it's GIF, I'm
39:20 out of
39:27 here. Is that how to do Vicki? Is that
39:30 how to do character consistency that
39:31 video? Yeah. Put it pop it into
39:33 irregulars.
39:42 D but once had something story of my
39:45 life. It can't be GIF. It better not be
39:50 GIF. It gonna be a short show. It's not
39:53 GIF. Brandon, you're wrong.
39:56 Wait. ChatBT will know.
40:08 Listen, we have to settle this once and
40:10 for all. How do you pronounce the word G
40:17 if the eternal civil war? The creator's
40:21 pronounce
40:23 pronunciation. Oh no.
40:31 Stephen White, the inventor of the
40:33 graphics interchange format, said it's
40:36 pronounced
40:37 GIF. Oh
40:39 [Laughter]
40:44 no, I heard it's GIF because generated
40:48 starts with a soft G.
40:52 Generated. But but it's not generated.
40:55 It's graphics. Graphics interchange
40:57 format. That's a hard G.
41:09 Udo and Udio. This is the udio thing.
41:13 Most of the internet, yeah, they say GIF
41:15 as in gift because, you know, it stands
41:18 for graphics and graphics has a hard G.
41:21 So, who's right? Technically, the
41:23 creator, practically the mob. TLDDR. GIF
41:26 is the creator's preference. Hard G is
41:29 the popular choice. my take. Say it
41:30 however you
41:31 [Laughter]
41:35 like. Oh, good lord.
41:43 Jesus. All right. This is what we're
41:46 going with. This This is the movie we're
41:48 making,
41:50 people. Sad
41:53 Kyle. This is like Well, no. We'll call
41:55 this carpet salesman Kyle.
42:07 I'm I'm totally confused about the
42:08 character consistency thing. I guess
42:10 maybe they mean character consistency if
42:12 the camera's moving around them, things
42:14 like
42:17 that. Um, I saved that. Let's go back
42:20 here. Let's trash that. We'll select an
42:23 image. We'll go grab an
42:25 image. Where did I save it? carpet
42:28 salesman
42:35 Kyle. Vicki, it's absolutely GIF with a
42:38 hard G. Thank you, Vicki. I feel
42:41 vindicated.
42:44 Uhuh. I'm having a tough time using my
42:47 new glasses. Wow, that's a
42:49 disappointment. We're going to have to
42:50 agree to disagree. I found this in
42:52 another video about MidJourney using the
42:54 same character.
42:56 Zappier is happier. Yeah, that I agree
42:59 with. Okay, we're going 16
43:02 by9. We're putting carpet salesman Kyle
43:06 in there. All right, like it.
43:11 Nice. Okay, so now we need a we need a
43:14 Let's see.
43:16 Um, do I still have
43:20 Yeah, a man stands. Okay, let's get rid
43:23 of all that [ __ ]
43:28 All right. Well, let's This
43:31 is It's going to do something here.
43:47 We we'll we'll put this in irregulars.
43:50 So So you can
43:53 Zeniad. Oh, I didn't even really look.
43:56 Did it use the correct logo? I think it
44:05 did. Yeah, it did. Irregular by
44:09 design. You know who you are. You talk
44:12 to AIS like old friends. You wear
44:14 glasses that double as a statement.
44:16 You've got 12 tabs open and one of them
44:19 is probably rewriting the laws of
44:21 perception. The world calls it weird.
44:24 Zeni eyewear isn't just about seeing
44:26 clearly. It's about being seen clearly
44:29 for tinker for the tinkerers, the
44:31 dreamers, the ones who build rockets in
44:33 their garages and host salons in and
44:37 then it cuts it off. That's perfect. I
44:39 love that it cuts it off.
44:47 That's pretty good. I like it. All
44:50 right. So, this is going to be in uh in
44:55 irregulars. 12 tabs. Yeah, that's
44:57 lightweight. Lightweight 12 tabs. I have
45:01 50 open right
45:05 now. Oh, what was that? That was Oh,
45:08 that was the Oh, from digital gods. It's
45:10 getting so good. Yeah, this was a cool a
45:12 cool ad here. Let me put on the sound
45:13 because this was a good
45:35 one. Why? Why am I having to do this?
45:41 What the [ __ ] was that?
45:50 Oh,
45:51 wait. No, you can't see it because it
45:54 [ __ ] spawned a new
45:59 [Music]
46:03 window. This is turning into meltdown
46:06 Monday. I can't be held responsible.
46:09 I
46:10 can't. But look at look how nice I look.
46:13 We're not weird. We're
46:16 irregular. I look like the Italian
46:19 Jewish the Italian Jewish third cousin
46:24 of
46:30 me. All right, let's go to YouTube.
46:39 [Music]
46:48 Share screen.
46:56 [Music]
47:10 All right, everybody. It looks like Hey,
47:11 what's up guys and welcome back to Ask
47:12 ENK. So, the folks at Runway, an applied
47:15 research company at the forefront of
47:16 artificial intelligence in terms of
47:18 media, have unveiled a brand new version
47:20 of their video generating tool called
47:22 Gen 4. And currently, this is the most
47:24 advanced AI model till date. Now we
47:26 already know that when you try to
47:27 generate multiple images or even
47:30 multiple videos, we know AI struggles in
47:32 terms of consistency and that is what
47:34 has been addressed with this new
47:36 release. As Gen 4 puts the idea of
47:38 non-consistency in terms of objects,
47:41 characters and also scenario, a thing of
47:43 the past. As Gen 4 now introduces the
47:45 concept of world consistency, enabling
47:48 the creation of coherent worlds,
47:50 objects, locations, and characters
47:52 across different scenes, allowing for a
47:54 more seamless depiction of the same
47:56 characters and settings in diverse
47:58 scenario. Gen 4 sets a completely new
48:00 standard for both image and video
48:02 generation. And according to the folks
48:04 at Runway, this is first model ever that
48:06 they've put together. Nah, this is just
48:08 gonna talk about it. There's no there's
48:10 no actual Oh, is there wait maybe there
48:13 is wooden toy which he used as a
48:16 reference and used another image that
48:19 was New York City to create a
48:20 composition. Another thing that was
48:21 thrown into the mix was to prompt for
48:23 characters to move in front of the toy.
48:25 Every single time characters
48:27 Why? Maybe I'm in the wrong interface.
48:29 Maybe that's why it's so [ __ ]
48:30 confusing.
48:33 But let's see. Let's see this video.
48:36 Wait, am I is
48:40 what? You pulled the
48:50 instructions. Oh, you can't see it.
48:52 Okay. I I am I am so done. I'm done. I'm
48:55 done. I'm done. I'm done.
49:04 Look at the This is so
49:12 sad. You look like Andy Garcia's little
49:22 brother. Is there an easy way to use the
49:24 platform for runway? The last time I
49:26 tried runway, it wasn't very friendly.
49:28 Well, that's the problem I'm having. I
49:30 just saw something in that
49:32 video. Get your notifications. Yeah,
49:35 that's fine. You can notify
49:38 me. But where? What? What am I missing?
49:40 and go to
49:47 dashboard act
49:50 one
49:52 prompt
49:55 camera expanded
49:58 video turbo gen
50:02 4 gen
50:07 4 what am I missing because over
50:10 in that YouTube thing. Is this
50:20 it? So, it says image references zero of
50:26 three. Maybe I need to log out and log
50:29 back
50:30 in. Hang on a
50:36 sec. Get started.
50:42 Generate
50:47 video. Drop
50:53 image. I know the sweater magically
50:56 appeared on the other thing. Where are
50:57 you flying off to? I'm trying to find
50:59 interface here. View all
51:04 sessions. This shouldn't be this hard.
51:07 This shouldn't be this hard. This
51:10 shouldn't be this
51:13 hard. Runway watch. That's just watching
51:17 other people's [ __ ] where they've
51:19 figured out how to use
51:25 it. Is it a screen redraw problem?
51:32 No. That's my
51:35 history. Any ideas?
51:39 Anyone? Runway New Level. Very
51:42 interesting. That's something. Steve,
51:44 what did Steve come up
51:46 with? Not all heroes wear capes, but
51:48 this one
51:50 does. Kyle's still going. I know.
51:53 Because it's pronounced GIF. I should be
51:54 able to just take
52:01 off. Gen 4.
52:07 Gen 3
52:10 camera. Oh, we see. Oh, damn that
52:22 thing presets.
52:28 Turbo Gen 3 Alpha Gen
52:34 4. View the guide. All right, Vicki. I'm
52:38 reading the
52:40 [Laughter]
52:48 manual. No, there's nothing there.
52:54 [Applause]
53:02 So, you're supposed to be able to put in
53:03 three images according to this video
53:06 that we were just on. Image
53:08 reference. Oh, when you're making an
53:11 image, there's an image
53:14 reference. Okay. Wait, wait for it.
53:18 Making an
53:20 image
53:29 frames
53:31 settings fixed
53:34 seed aesthetic
53:37 range. What the [ __ ]
53:41 References. No style references.
53:47 anyone first. Don't underestimate the
53:51 simplicity. My Gen 4 says soon. Yeah,
53:54 but I'm in the CPP. I should I I have
53:56 access to Gen
53:59 4. Are we having fun yet? Wolfman clan.
54:03 I
54:08 am Welcome to the AI learning lab where
54:11 the teacher is often the
54:13 [Laughter]
54:17 student. I'm [ __ ] done already. I'm
54:20 done. I'm done. I got I got nothing for
54:24 you people. Oh my god.
54:36 This shouldn't be this
54:39 hard. This should not be this hard. This
54:43 should not be this
54:45 hard. All right, hang on a sec.
54:49 [Music]
55:09 We can go make
55:10 songs. Drag them in from your
55:16 finder. Drag them to
55:19 where? Type a
55:22 prompt or take an image. So if I
55:26 go
55:28 finder carpet salesman
55:45 Kyle
55:49 H. So, let's grab dad and champ. And
55:53 then we'll grab Joker and drag both of
55:55 those
55:57 in. Nope. Won't let me drag in
56:02 multiple. Watch the video at 127. That's
56:06 where I was. Yeah, 127. This is what I'm
56:09 saying. See down here at the bottom, it
56:11 says image references zero of
56:15 three. And I'm in an image tab here.
56:20 The video lies then.
56:23 Yes. Apparently
56:58 It It must just be a thing where they
57:00 they don't quite have this fully rolled
57:02 out
57:03 yet. I don't get it. Hey, Kyle. Yeah.
57:08 Uh, if you go back to the screen you
57:10 were just on for images. Mhm.
57:14 No, that's not where you just were.
57:19 Uh, wait here. Yeah. Um, go back to the
57:24 image tab. Mhm. There is a
57:27 references subtab under style. It's not
57:30 a tab.
57:35 I've clicked it many times.
57:38 That's I thought that was the indicating
57:40 that that was the style tab and this was
57:41 the reference tab, but it's not a
57:45 tab. There's What's this? There's a
57:48 little paint
57:50 bucket
57:53 style. Wait. Oh, pull down menu. No,
57:56 that just made it go away. All my
57:59 styles.
58:01 Oh,
58:03 no. It says
58:05 soon. references soon. Okay. All
58:11 right. So [ __ ] done. All right. Thank
58:13 you, producer Brandon. I could have read
58:16 the word soon, but I have new glasses. I
58:18 can't see
58:20 anything. I'm going back to VO. I'm
58:23 going back to uh Lumal Labs, I
58:27 guess. Oh, this is maddening. All right.
58:30 Apparently, it's much better.
58:35 Um, I don't know. Let's let's let me let
58:38 me let me just take a [ __ ]
58:48 breather. If you want to find more
58:50 people that are as as awesome as I
58:56 am, go to the AI salon. Go to the
58:59 salon.ai AI and click join our community
59:03 and then go join our community. It
59:05 doesn't cost anything and it's it's
59:07 awesome and it's great people and you'll
59:09 learn a lot. Um you'll learn from people
59:13 unlike me who actually know what they're
59:18 doing. I listen here's the here's the
59:21 thing. I do feel better. I do feel
59:25 better
59:26 that the feature wasn't there and I just
59:29 missed the teeny tiny little soon
59:34 button. No. No.
59:43 No. All right. Who Anybody have any
59:46 questions about AI?
59:51 I I have to do a physical
59:53 dexterity
59:58 challenge. Um, tomorrow night we have a
1:00:02 uh is AI Salon presents and we've got
1:00:05 Brian Moahan and
1:00:08 Adnan
1:00:10 Eear Ectear. I I forget how to say his
1:00:14 last name. Adnan. Lovely man with a
1:00:17 challenging last name. um they wrote the
1:00:20 book um AI culture shift and uh they're
1:00:25 going to be speaking about their book
1:00:26 and how they produced it and how they
1:00:28 met. Ifar that's right
1:00:32 ifar if right if is where the the lead
1:00:37 pronunciation is.
1:00:45 Um, here's what I will promise you. I
1:00:49 will, now that I
1:00:51 have smashed my head against the wall
1:00:54 for what was that 45 minutes, half an
1:00:56 hour, um, trying to figure out character
1:00:59 consistency and runway, I will figure it
1:01:01 out. And once I figure it out, I'll show
1:01:02 you. But I will not go back there until
1:01:04 I know what the [ __ ] I'm doing because
1:01:05 that's that was bad. That was bad,
1:01:08 people. Bad Kyle on Sora you can create
1:01:10 two images at once with plus and four
1:01:13 with pro. Cool. Oh nice. Okay. So on
1:01:17 Sora instead of just doing one image in
1:01:21 chat
1:01:25 GPT not now
1:01:30 images. Oh yeah. Four up. Nice.
1:01:38 16
1:01:40 by9 upload from
1:01:43 device carpet salesman Kyle. No, I want
1:01:46 I want I want my real screenshot.
1:01:49 Where's my real screenshot? Oh, it's
1:01:51 over on screenshot
1:02:01 desktop. All right. So let's say
1:02:08 um be
1:02:11 spectacled be spectacled
1:02:23 superhero
1:02:25 go add it to the queue images. So, we're
1:02:29 making
1:02:39 images. I think uh Sam Alman said today,
1:02:44 um, so remember when chat GPT first came
1:02:48 out and they didn't think anyone was
1:02:50 going to use it and then it got to a
1:02:51 million users in 5 days? Well, this new
1:02:54 image gen feature, they got to a million
1:02:58 users in an hour. Oh, you can't see
1:03:00 anything,
1:03:02 [Laughter]
1:03:05 right? Of
1:03:08 course.
1:03:09 Um, yeah. So, tomorrow is the AI
1:03:15 um AI salon presents and then Wednesday,
1:03:18 Sabrina Ramanov is presenting Blot
1:03:22 uh in the AI
1:03:24 salon. And then AI life hacks is also
1:03:27 Wednesday. Get your ass in the AI salon.
1:03:29 If you're not in the AI salon, what are
1:03:31 you doing with your life? I got to tell
1:03:32 you, I just I just wrote a piece about
1:03:34 this.
1:03:36 Um where did I write it? I think I wrote
1:03:38 it on LinkedIn.
1:03:40 I might have wrote it on on X. I don't
1:03:42 remember. But
1:03:45 um I think we're in a place where it's
1:03:48 no longer optional that you get AI
1:03:50 literate. And I know I'm preaching to
1:03:52 the crowd, preaching to the choir here,
1:03:56 but describe your
1:03:59 video. I'm not in
1:04:04 videos. You got to be kidding me.
1:04:11 images. Did
1:04:13 it Did it not take
1:04:16 my activity? Oh, here we go. Okay.
1:04:28 Jesus. I can't I can no longer I can
1:04:32 listen. I cannot be
1:04:36 respon. Those are
1:04:45 videos. Why is it making me older? I'm
1:04:48 not that
1:04:49 old. And why did it
1:04:53 I What did I
1:04:55 do? Solomon wants to know the best AI
1:04:58 for voice. Okay.
1:05:04 Um, what is it?
1:05:06 Cartisia.ai. Cartisia.ai.
1:05:12 Yes. Oh, cool. Saying hi from Kenya.
1:05:16 What's happening, dude? Uh, apologies to
1:05:19 anyone that's new here tonight. This
1:05:22 channel, while it is special and awful,
1:05:25 is not usually this awful. I don't know
1:05:28 what's going on.
1:05:31 Um, okay. There there's two places if
1:05:34 you want to do voice. There's there's
1:05:36 kind of the the the old school king of
1:05:39 the hill on voice is
1:05:47 11labs
1:05:59 11labs.io Jeff Flanigan. You've aged in
1:06:02 the last 30 minutes, Kyle. I have, man.
1:06:05 I like I just I just want [ __ ] to work.
1:06:08 It's all I I don't want much. I just I
1:06:11 don't I don't want
1:06:13 much. I just want it to
1:06:20 work. Engineering teams. Hey Kyle,
1:06:23 what's the acceptance
1:06:24 criteria? The acceptance criteria, it
1:06:28 should [ __ ] work.
1:06:35 And then the other one is um Cartisia K
1:06:40 K A R T E S I
1:06:44 A
1:06:50 Cartisia.ai. Oh, no. It's play. It's
1:06:54 play.cartisia.ai. play.cartisia.ai.
1:07:05 And this is a site I'd never heard of
1:07:08 before and it's really
1:07:12 good. It's very much like in design. You
1:07:15 can design a voice. You can just
1:07:17 describe
1:07:18 it and it'll make it for you. Um, and
1:07:21 it's got good emotion. The other one you
1:07:23 can try is Hume. H U M E. What's What
1:07:27 was Hume? Hume is Hume.
1:07:30 Uh, oh, by the way, if you're a
1:07:33 developer, Hume.AI, if you're a
1:07:36 developer and you're wondering, should I
1:07:39 spend the extra money for
1:07:41 do the answer is
1:07:46 yes. I don't care if you
1:07:51 use.ai.xyz.upyoub butt dot. It does. No
1:07:54 one can remember your stupid [ __ ]
1:07:56 domain unless
1:07:59 it's.com. Now, you might be saying, "But
1:08:01 Kyle, you're old and tired and you had
1:08:05 one of the early dot agencies back in
1:08:09 the 90s, and you're
1:08:11 bitter. You're damn right I'm
1:08:14 bitter. No one can remember your stupid
1:08:16 [ __ ] domain." All
1:08:21 right. I can't see my pen, please. Which
1:08:26 pen? Um, Kyle, did you tell the nice
1:08:29 people? Oh, about our conversation with
1:08:32 Kay, coming up. I did not. Oh, that's
1:08:34 the other thing. So, yesterday we've got
1:08:38 we we we made this thing, Ann and I,
1:08:41 we've got the new podcast, the AI
1:08:42 readiness project. So, Wednesday at 4
1:08:46 PM, that episode's going
1:08:49 live on the AI Salon YouTube channel,
1:08:53 the Sheile Leads AI YouTube channel, and
1:08:56 probably a fair amount of other places.
1:08:58 All right, so that's happening, too. And
1:09:02 I know if you're like, "Is it as much
1:09:04 fun as this show is?" Oh, it's more fun
1:09:06 than
1:09:08 that. There's other people there that
1:09:10 that can bail me out if I flail.
1:09:13 [Laughter]
1:09:20 I just I I don't know what to tell you.
1:09:22 I don't know what to tell you. There was
1:09:23 a time when this
1:09:25 show had it going
1:09:33 on while we're flailing.
1:09:36 Higsfield.ai. Is this Is this going to
1:09:38 be some Microsoft
1:09:40 product? Higsfield. Is this Higsfield
1:09:44 Bosen?
1:09:47 Higsfield.ai. Are we going to get to
1:09:48 spin up a large
1:09:52 collider? Motion
1:09:58 controls. What is
1:10:00 this? What is
1:10:03 this? Oh, this is just a prompt thing,
1:10:06 isn't it?
1:10:16 Dolly
1:10:18 in image to video. Text to
1:10:28 video crash. Zoom out.
1:10:38 This is a prompt
1:10:41 library. What? Try real
1:10:48 magic. What is this thing? So from what
1:10:51 I'm seeing is if you can add an image
1:10:55 and add a prompt to do motion control.
1:11:00 For example, you want it to do a 360
1:11:03 camera spin around the subject in the
1:11:06 picture and it will generate the camera
1:11:09 control around the subject while
1:11:13 maintaining consistency. Yeah. But
1:11:15 aren't they just using this is just a
1:11:17 prompt. This is a wrapper sitting on top
1:11:19 of an existing IM uh video gen tool,
1:11:23 right? Because you can do this now with
1:11:26 Luma's got camera control. Runway's got
1:11:28 camera control. I think there Clling has
1:11:31 camera control. I think there's three or
1:11:32 four of them that have it. Do you know?
1:11:35 I don't know if it's a rapper. It feels
1:11:39 like they are doing their own thing. I
1:11:42 don't buy it. Came to us via Twitter. I
1:11:46 don't buy
1:11:47 it. I don't buy it. Wait, where can we
1:11:51 try real magic login? Sign
1:11:55 up. What's this? Is this the
1:11:58 homepage?
1:12:00 Pricing. I bet pricing is going to be a
1:12:03 lot. Oh, it's not too
1:12:04 bad. Pick your plan. 70 video
1:12:07 generations. 150 video
1:12:10 generations monthly. Well, it's not
1:12:13 great. 80 bucks a month for
1:12:17 150 [ __ ]
1:12:21 video. All right. Well, it's pretty cool
1:12:24 if you if you don't know camera stuff.
1:12:26 So, what is this? Higsfield.ai.
1:12:30 If you don't know, uh they are
1:12:31 partnering with Google Cloud to
1:12:34 accelerate democratization of social
1:12:36 media video creation with AI. Um they
1:12:39 are also on hugging face. So, I do
1:12:42 believe that this is a standalone model.
1:12:45 Well, then this is good interface
1:12:47 because what the other what the what the
1:12:49 other um the reason I'm thinking this is
1:12:51 a rapper is that most of
1:12:54 the video companies are putting it all
1:12:57 on you to understand what camera
1:12:59 controls are and what what moves to use.
1:13:03 And so the fact that they've kind of
1:13:05 pre-prompted their own model pretty
1:13:08 smart through object out generate. Yeah,
1:13:12 it's pretty cool. I figured you could uh
1:13:15 throw Kyle uh with the glasses on there
1:13:19 to see if it could do any better than
1:13:21 what we didn't what we weren't able to
1:13:23 produce with uh runway since you didn't
1:13:26 have it. Okay. All right. So, let's do
1:13:30 image to video. There's image to video
1:13:35 um through object. Okay. So, so we'll
1:13:37 say a
1:13:41 world of
1:13:47 confusion is inside
1:13:51 his left
1:13:53 [Laughter]
1:13:58 eye. I got to sign in. Oh, I probably
1:14:01 don't have
1:14:04 a I probably don't have an account, so I
1:14:08 probably can't make
1:14:15 any. Is it going to let me do
1:14:18 it? I lost my
1:14:21 prompt. Come on. Come
1:14:25 on. Really?
1:14:30 Mouth in. All right, let's do mouth in.
1:14:33 Generate image to
1:14:38 video.
1:14:41 There's bugs in his
1:14:44 mouth.
1:14:46 Generate. It's going to take me. Upload
1:14:49 the image. I did upload the image.
1:15:04 Janky is as janky does. Janky jank jank
1:15:09 jank. Generate.
1:15:13 If this works, it kind of feels like
1:15:15 what Pika is doing with Pika effects.
1:15:18 Yeah, exactly.
1:15:25 There's good news and bad news about all
1:15:27 these video tools in my
1:15:31 opinion. The good news
1:15:35 is you don't need to buy a camera
1:15:37 anymore, right? We're going to be able
1:15:39 to just generate [ __ ] You can buy a
1:15:41 camera. You can go shoot some acting
1:15:43 performances if you want, but you can
1:15:44 probably just now start to do acting
1:15:47 performances just by recording audio or
1:15:50 a face with a with an iPhone and then
1:15:52 just generate the rest. Um,
1:15:55 but I think this is going to be some
1:15:58 amount of
1:16:00 years, two, three
1:16:02 years before generating good quality
1:16:05 video storytelling. you know, multiple
1:16:08 scenes, sound effects, audio, background
1:16:11 music, you know, color grading, [ __ ]
1:16:14 like that. Like all that stuff, I think,
1:16:16 is going to remain really complicated
1:16:18 for some time. So, if you're not scared
1:16:21 off scared off by complexity, then video
1:16:25 is going to be fine. If complexity wigs
1:16:28 you out, I wouldn't play with video
1:16:30 tools right now other than do a little
1:16:33 scene here or there that are cute. go to
1:16:35 peak effects
1:16:36 and peel a car off off a calendar or
1:16:43 whatever.
1:16:46 Um the knowing where to go for which
1:16:50 tiny little thing is better. There are
1:16:53 like 15 different video tools. Some of
1:16:57 them are better at other things. It's
1:16:58 it's almost impossible to to know. And
1:17:01 then they change a lot. They change a
1:17:05 ridiculous amount. Um, did we do
1:17:08 anything over No, we last thing we did
1:17:10 was the Zeniad over here. Did we do
1:17:14 anything on runway? No, we did sad fat
1:17:19 aging Kyle. It's
1:17:26 depressing. And then where was I just
1:17:28 what were we in? Oh, the real magic one,
1:17:31 right?
1:17:38 Yeah,
1:17:39 Higsfield.ai. All right, there you go.
1:17:43 Generating takes generating bugs takes
1:17:51 time. Um, the other thing that I did
1:17:54 want to we'll we'll let this thing
1:17:56 generate. We'll come back to this. The
1:17:57 the other thing I did want to talk
1:18:00 about is I want to go to
1:18:04 X and something something hit me today
1:18:08 with the the way
1:18:10 [Music]
1:18:17 um with the way the image generation
1:18:20 tool is
1:18:24 working in chat GPT.
1:18:35 Hey, let's
1:18:38 see.
1:18:40 Um, social ad. Oh, no. Let's do uh let's
1:18:45 do um
1:18:50 interface mockup.
1:19:20 Okay. So, I'm going to take a screenshot
1:19:23 of this. So, somebody did
1:19:25 this F1
1:19:30 editorial kind of cool thing, right?
1:19:36 So, I'm just going to screenshot that.
1:19:37 I'm going to go into chat
1:19:39 GPT and then I'm going to upload that
1:19:44 thing and I'm going to say
1:19:54 um make me
1:19:57 a
1:20:00 website mockup for an for a
1:20:11 photographer
1:20:13 website that is inspired
1:20:18 by these
1:20:29 ads. Oh, wait.
1:20:33 Stop. I don't
1:20:37 want the
1:20:40 code. I
1:20:44 want a visual of
1:20:48 what the final will look like. Although
1:20:52 I probably should have let it make the
1:20:54 code because it probably would have
1:20:55 coded it
1:20:57 too. But you can take any
1:21:00 style and just say, you know, put this
1:21:03 thing into that style.
1:21:07 Or you can take any photograph and say,
1:21:10 "Turn this into a Lego scene and it'll
1:21:14 do it and it knows what that is." Or,
1:21:16 "Turn this into an Atari 2600 game."
1:21:20 Like, it's got enough knowledge of the
1:21:22 world that you can just ask it weird
1:21:23 [ __ ] and it it can interpret what your
1:21:27 image is.
1:21:29 And it's bizarre. It's wild. It's wild,
1:21:34 people. It's
1:21:36 wild. All right. Higsfield AI. If you're
1:21:39 not paying for it, it's slow. All this
1:21:42 stuff is
1:22:00 slow. One more reason to watch this on
1:22:03 YouTube. I can review right through
1:22:08 this. Oh, right. Tomorrow's April 1st.
1:22:11 Good luck with
1:22:15 that. Wonder what I'm going to do. I
1:22:17 might actually teach you
1:22:18 [Laughter]
1:22:28 something. Send a screenshot of mods for
1:22:31 images and Sora. Cool. Thank
1:22:36 [Music]
1:22:45 you. Okay,
1:22:48 cool. Thank
1:22:52 you. These crazy interfaces,
1:22:56 man. So, if you Oh, you go here. Oh, I
1:22:59 see. Images. Got it. Okay. Yeah, that's
1:23:03 completely not clear. It is. Once you
1:23:06 know it's there, it's
1:23:09 clear, but I'm pretty sure that I
1:23:14 did. And you can't do images in 16
1:23:18 by9. So, you're going to
1:23:21 start I I just can't
1:23:25 I Somebody needs a cranky old Gen Xer in
1:23:28 the corner.
1:23:39 Bugs in the
1:23:43 mouth. That's disturbing.
1:23:51 [Laughter]
1:23:59 Oh
1:24:03 man, was that worth 10
1:24:06 minutes? My teeth look
1:24:17 nice. Oh, this is the metaphor of
1:24:19 tonight's show. Yeah, exactly. starts
1:24:22 out like this and then all of a sudden
1:24:24 you're like, "Wait, what? What? What?
1:24:26 Why am I
1:24:28 here?" Okay, I'm putting away Higsfield
1:24:31 AI. I don't care about
1:24:33 it. I'm I Everything's too complicated.
1:24:37 Everything's too complic Oh, look how
1:24:38 cool that
1:24:41 is. Remember why you
1:24:45 shoot? That's pretty slick. Let me try
1:24:48 one other thing.
1:24:50 New plus. We're going to upload. We're
1:24:53 going to upload a scene from Yan
1:25:16 computer. Calm down. I'll be with you
1:25:18 people in a second.
1:25:21 What's he doing? This is very
1:25:31 annoying. Okay, this is good. This is
1:25:33 good. All right, so we're going to
1:25:34 upload this
1:25:36 picture of two boys and a dog on a
1:25:39 beach. And I'm going to say, "Turn this
1:25:43 into a Lego kit.
1:26:05 called.
1:26:08 Um, summer at the
1:26:14 beach, a Lego kit
1:26:21 box. I was going to say something very
1:26:25 inappropriate for summer at the beach,
1:26:28 and I chose not to. You're welcome.
1:26:32 [Laughter]
1:26:38 Uh, just one more thing. Hold, please.
1:26:41 It's like, shouldn't he just end the
1:26:44 show? I listen, I called it when I said
1:26:47 if the creator of the the GIF calls it
1:26:52 GIF, I'm going to end the show. And what
1:26:54 didn't I do? I didn't end the show. So,
1:26:58 when you operate in this world without
1:27:00 integrity, what happens? You make an ass
1:27:03 out of the guy that [ __ ] called GIFs
1:27:09 GIF. That's his problem. It's not my
1:27:15 problem. Oh, another thing while we're
1:27:18 waiting for that to
1:27:21 render. Um, US
1:27:24 patent on toilet
1:27:27 paper. You're like, where's this
1:27:30 going? Oh, I'll show
1:27:40 you. No. Where's the
1:27:49 patent? Patent images. Here we
1:27:54 go.
1:27:56 Okay. There you
1:27:59 go. Controversy
1:28:02 solved on the patent. It's
1:28:06 over. It's over. It's not under. It's
1:28:11 over. Anyone comes at you with the under
1:28:15 theory, you pull out the
1:28:18 patent.
1:28:21 Okay. Clearly, this is going down the
1:28:24 toilet. This is going to be great. Just
1:28:27 one more thing. Oh, please.
1:28:29 [Laughter]
1:28:34 Mr. It has chat GPT listened to my live.
1:28:37 That's not a smart
1:28:39 [Laughter]
1:28:43 move. Wait, something went
1:28:46 wrong. Oh, come on. Try again.
1:28:56 I this has been crap in the bed all day,
1:28:58 all night. It
1:29:00 like open AI servers. This must be
1:29:03 really
1:29:04 ugly is all I can
1:29:08 say. That's the only way. And I will
1:29:11 change yours if it's under. Exactly,
1:29:13 Vicki. That's the other thing. If you're
1:29:16 at someone's house and they have they
1:29:17 have the toilet paper under and and you
1:29:20 have you're fighting for it, fighting to
1:29:22 find it. You take that [ __ ] off there
1:29:26 and you flip it
1:29:28 around. Is it passive aggressive? Yeah.
1:29:31 Do you need Do you talk about it? No.
1:29:34 You just correct it quietly. You let
1:29:36 people experience the joy of the
1:29:40 over the over roll.
1:29:45 You just go back and you continue on
1:29:47 with your your afterdinner mints and
1:29:50 your cocktail. You don't say a thing.
1:29:53 You don't say a thing. And then if
1:29:56 someone comes out, did anybody change
1:29:57 the toilet paper? You're like, "Huh? No.
1:30:01 It's how it was when I was in
1:30:03 there." Totally passive aggressive. It's
1:30:06 the one time in life being passive
1:30:09 aggressive is actually approved by the
1:30:12 passive
1:30:13 aggressive
1:30:15 committee on
1:30:27 appropriateness. This doesn't look
1:30:29 photographic. You you dumb
1:30:33 dumbs. Or does it? Maybe it does. Maybe
1:30:36 up close it
1:30:37 does. Toilet paper math is the hardest
1:30:41 math. I don't like places that lock
1:30:43 their TP and you can't change it. Who
1:30:45 does that? What kind of monsters do
1:30:48 that? Please use as a style or
1:30:52 source. Oh, this is the other thing.
1:30:54 Okay, I'm done with chat chat GPT.
1:30:58 Now they are they're just they're doing
1:31:02 this thing where if they think they're
1:31:04 going to get sued, they generate the
1:31:06 image and then delete
1:31:08 it. All right. I I can't do this
1:31:10 anymore. I'm leaving. I'm leaving. I'm
1:31:13 done. I'm done. Okay. Here's the deal.
1:31:16 Yesterday, no, Friday, I said that
1:31:19 everything changed. You got to go all in
1:31:20 on AI. Everything's right with the
1:31:22 world. We've hit AGI because this
1:31:24 picture maker thing is so good. Today
1:31:27 I'm going on record. Nothing in AI
1:31:29 works. Give up. Don't just go back to
1:31:32 doing normal lite work like like use
1:31:37 spreadsheets where you actually have to
1:31:38 go in and type in formulas and use a use
1:31:43 a word processor where oh the keyboard
1:31:46 on your computer used to be used for you
1:31:48 go into a word processor and you would
1:31:50 actually write content instead of going
1:31:53 into the chat hole and and talking a
1:31:57 word document into existence in the
1:31:59 before time you actually use your
1:32:01 keyboard to write content. I know it
1:32:05 sounds weird, but you would like write
1:32:06 an article. Or if you wanted to write a
1:32:08 book, it would be like eight months if
1:32:11 you're fast or like 10 years and a lot
1:32:14 of whiskey if if you're
1:32:16 not. So, so go back to that. Don't AI is
1:32:22 stupid. I I don't I should have never
1:32:24 started the channel. I I I thought it
1:32:27 was going to be good and then like as of
1:32:30 Friday it was and then today like three
1:32:34 days
1:32:36 later
1:32:38 just if there's like a workshop on on on
1:32:42 plumbing repair, I I'd take
1:32:46 that. I'd get into plumbing. Um
1:32:50 uh car repair. Uh gardening could be
1:32:54 good. um guitar maintenance because
1:32:58 there's gonna be a lot of people out of
1:32:59 work singing songs and their guitars are
1:33:01 going to be out of tune and [ __ ] up.
1:33:03 So you
1:33:04 can But I would just no AI. Don't use
1:33:08 AI. As Brandon said, there's an AI for
1:33:11 that.
1:33:14 Exactly. The chat hole snort. All right.
1:33:19 Uh HT Snow Day said, "I I might be
1:33:21 slightly overreacting and and that AI is
1:33:24 probably going to be fine, but but
1:33:27 apparently not
1:33:30 tonight. Apparently not tonight. I can't
1:33:33 do
1:33:34 anything. I'm watching a video. Here's
1:33:37 how to do the thing in your [ __ ]
1:33:38 thing. My interface doesn't look like
1:33:40 that." And they put a word. I don't
1:33:43 Here's here's the equivalent. Let's see.
1:33:46 Wait. Okay.
1:33:51 Here's the interface elements, right?
1:33:54 This box, the interface elements are
1:33:56 that big. This little teeny tiny thing
1:33:59 is the word that says soon. So, here's
1:34:02 the button that I clicked on that
1:34:05 says image whatever influences up
1:34:09 uploads. And and here was the little
1:34:11 button that said soon.
1:34:19 I want to smack
1:34:20 someone. I want to smack
1:34:23 someone. Whoever Whoever did that, I
1:34:26 guarantee you that was a little
1:34:27 22year-old [ __ ] with actual functional
1:34:31 eyeballs. I'm sure they can see it.
1:34:35 Well, uh, you know, if I if I wanted to
1:34:37 to actually put the word soon in there
1:34:39 so it was legible, it would be too the
1:34:42 column the CSS stylesheet for the column
1:34:44 would exceed the boundary of the so so I
1:34:47 made it in threepoint
1:34:51 type. Way to go,
1:34:54 Skippy. Way to
1:34:57 go. Way to make old people feel bad.
1:35:10 April fools. Soon
1:35:13 soon. April fools. The 22year-old that
1:35:16 made that interface element's going to
1:35:18 get a
1:35:19 [Laughter]
1:35:21 promotion. Little
1:35:25 [ __ ] Okay. Uh, give up on AI
1:35:28 but come back tomorrow. Yeah, you might
1:35:29 as well come back tomorrow. So, tomorrow
1:35:31 we're going to talk about um uh
1:35:33 gardening, planting seeds. Which seeds
1:35:35 do you
1:35:36 plant? Um I'm going to start selling
1:35:38 Madagascar vanilla beans, and I will be
1:35:41 giving twerking lessons. So, we're going
1:35:43 a little more old school. We're going a
1:35:44 little more analog uh in person. Uh
1:35:48 we're not going to be doing a AI
1:35:50 anymore. We'll come up with some other
1:35:51 two words that AI stands for. So, it'll
1:35:54 still be the AI learning lab. It'll be
1:35:56 like all interference or, you know,
1:35:59 we'll come up with something. And then
1:36:01 we're just going to do gardening and and
1:36:02 we're going to sell things from the Tik
1:36:04 Tok shop and I'll teach
1:36:06 twerking. I might do custom maninis, you
1:36:10 know, because I can model them and I
1:36:13 think that would be good for
1:36:15 everyone. Wait a minute. We could do
1:36:17 print ondemand maninis. Well, but we
1:36:20 can't use AI. So, we'll Okay, so we'll
1:36:22 go to stock image sites. We'll license
1:36:26 stock photography and stock
1:36:28 illustrations and then we'll allow
1:36:30 people to choose their design and they
1:36:33 can position it on the manini and then
1:36:35 it'll it'll be my body with the manini
1:36:38 and then they can position their image
1:36:40 on it. Okay, this is good. This is good.
1:36:43 We're gonna have to write the code from
1:36:45 scratch because vibe coding doesn't work
1:36:46 anymore because AI is stupid.
1:36:49 So we So we're going to
1:36:53 learn
1:36:54 JavaScript
1:36:56 and
1:36:58 e-commerce and
1:37:00 HTML and
1:37:03 CSS. So so we're going to start learning
1:37:05 that. So over the next six months I'm
1:37:08 going to learn CSS and then we'll do
1:37:11 JavaScript after that. So, we will
1:37:12 launch the Manini customization site in
1:37:17 2027 February, something like that. We
1:37:20 we'll do a plan and then uh and then in
1:37:23 between we'll talk about gardening and
1:37:26 uh and then I'll try to sell you some
1:37:27 Madagascar vanilla beans. All right.
1:37:30 Hope that works out for you. Hope that's
1:37:36 good. Unsubscribe.
1:37:42 Oh no, I'm on vacation next
1:37:48 year. Antique Ironing Lab. There you go.
1:37:51 Yeah, it's the antique ironing learning
1:37:53 lab. Exactly. That we will we will learn
1:37:57 fabrics and weaves. We will learn
1:38:00 stitches. We'll learn crocheting macra.
1:38:03 It's going to be fantastic survival
1:38:05 skills.
1:38:08 you know, vacuum seal your vanilla beans
1:38:11 because you don't want them molding in
1:38:12 the
1:38:14 shelter. You
1:38:19 know, I' I've had it. I have clearly had
1:38:23 a mental breakdown tonight.
1:38:26 Fine. It's fine. Fine by
1:38:33 me. All right, then. Bye. Um, tomorrow
1:38:37 AI Salon uh presents is at 5:00 PM
1:38:41 Mountain. Go to the
1:38:44 salon.ai. This URL right here. Go
1:38:48 there and click on join our community if
1:38:52 you haven't done it
1:38:53 yet. And then
1:38:59 uh Pate wants to know what he missed.
1:39:04 Pate, I know that your expectations are
1:39:07 already low for this channel.
1:39:12 Um, it is hard for me to describe to
1:39:16 describe how bad how nothing you
1:39:20 missed this episode. Like it was
1:39:23 staggeringly nothing. Like I know most
1:39:26 times it's like nothing and then a
1:39:29 little bit. Th this was like nothing and
1:39:32 minus a lot. So So So you not being
1:39:37 here, you're net
1:39:39 positive for for what you didn't
1:39:42 [Laughter]
1:39:47 miss. All right, I'm leaving. I don't
1:39:51 like this anymore. I don't like my job.
1:39:54 I don't like my job that I gave
1:39:57 myself. This is dumb. AI is dumb. Don't
1:40:01 do AI. If anyone tells you AI is going
1:40:03 to be big, just slap them. Go, No, it's
1:40:06 not. It's dumb. I heard it last night. I
1:40:08 learned it on the AI learning
1:40:10 lab. All right. Bye.
1:40:16 I just hung up on