8/27/2025 - Nano Banana to Gemini 2.5: Exploring Google's Image Generation

Description

Chapters

Transcript

0:05 [Music]
0:19 [Applause]
0:21 [Music]
0:36 [Music]
0:41 What is this?
0:46 [Music]
0:56 Woohoo!
1:08 [Music]
1:09 Woohoo!
1:15 [Music]
1:22 [Music]
1:39 Hello.
1:43 Hello.
1:48 [Music]
1:59 Is this place I can rest my forhead,
2:06 gather my thoughts in sweet silence
2:13 The bus is place where the feelings
2:16 aren't dead.
2:18 [Music]
2:20 From over exposed to balance
2:24 is this place I can slowly face. The
2:28 only one I truly can know.
2:32 These are tears from a long time ago.
2:36 Got these tears from a long time ago.
2:40 I need to cry. 30 years or so.
2:45 These are tears from a long time
2:51 ago.
2:56 [Music]
3:00 Oh darling, oh darling, say unto me,
3:07 where have you been? All my lifetime
3:11 [Music]
3:15 I have been swimming seven sad seas.
3:19 [Music]
3:21 Four women have to cost me their
3:23 lifelines.
3:25 I'd pull them into the waters in I'd
3:29 have warned them but I didn't know
3:34 these are tears from a long time ago.
3:38 Got these tears from a long time ago. I
3:42 need to cry 30 years or so.
3:46 These are tears from a long time.
3:52 No.
3:59 [Music]
4:10 Uh, what's happening good people? What's
4:11 shaking? What's going down?
4:14 Happy Thursday night. It's Thursday
4:16 night.
4:18 Probably going to do some more Nano
4:19 Banana.
4:22 Why are we focusing on this Nano Banana?
4:25 Why did Google change the name from Nano
4:28 Banana to Gemini 2.5 Flash image
4:33 generation module? Really?
4:37 Everyone's talking about Nano Banana.
4:43 Well, yes. uh be being a a responsible
4:47 uh corporate upstanding citizen. Uh we
4:49 we named uh the Gemini 2.5 Flash is it's
4:54 consistent with the well you've got G
4:55 Gemini 2.5 Pro which is the the larger
4:58 model. Uh it's of course it's a
4:59 multimodel mole model uh which it means
5:03 of course you know it's got different
5:04 modes uh of media that that that it can
5:07 process natively. One of those of course
5:09 is image. Uh and then in in the 2.5
5:12 flash which is the smaller model uh the
5:14 the uh the uh image generation tool is
5:17 uh great greatly improved and uh the fid
5:20 fidelity and promp prompt adherence and
5:22 and coherence uh quite frankly is uh is
5:26 considerably improved.
5:29 Nano banana. Yeah. No, that was the code
5:31 name that was never intended to be uh
5:34 publicly uh disclosed or correlated with
5:38 Gemini 2.5 flesh, which is the official
5:40 name, just to be clear. It's not it's
5:42 not Nano Banana is not the
5:45 I I know people seem to like it and it
5:48 and we we did make banana images. So, so
5:50 there are banana
5:52 but yeah it's called it's called Gemini
5:54 2 2.5 flash with it's it's the image
5:57 generation within within native natively
6:00 generating within really
6:05 [Music]
6:12 [Applause]
6:18 [Music]
6:22 Yeah.
6:24 [Music]
6:37 Heat.
6:43 [Music]
6:51 [Applause]
6:54 [Music]
7:02 Heat. Hey. Hey. Hey.
7:06 [Music]
7:21 Um,
7:23 what else been going on?
7:26 Um
7:29 there was I think probably worth talking
7:31 about
7:33 um I don't want to give it too much
7:35 credence but I think it's important to
7:37 note that that like one of the things
7:40 that
7:41 I talked about I don't know six eight
7:44 months ago is that I think that as as AI
7:47 capabilities get stronger that the
7:49 resistance to AI is is also going to get
7:52 stronger and there was a uh an AI I
7:55 meeting in Portland. I think it was last
7:57 night. And while they were in having the
8:00 meeting, a bunch a bunch of their cars
8:01 got vandalized and there was anti- AI
8:04 crap spray painted on the cars. One of
8:07 the things that we talk about in the AI
8:09 salon all the time, like one of our
8:11 values is empathy. Um, and especially
8:15 when people do stuff like that that's
8:18 that's violent and, you know, and
8:21 aggressive. Um I think it's important
8:25 it's important to understand that people
8:27 are scared and you know the anger comes
8:29 from fear and you know I think it's um
8:38 I think it's that fear that you can't
8:40 stop this thing right and so so they're
8:44 trying otherwise to stop it um and
8:46 that's kind of the the whole purpose of
8:48 this channel is that you know technology
8:51 ology has never progress has never not
8:54 progressed, right? Technology
8:55 progresses. It always has. When the
8:59 steam loom when the steam powered loom
9:02 showed up, you know, there was there was
9:05 all sorts of riots and things like that.
9:07 Tik Tok camp. Sorry about that.
9:10 Um Oh, that was about a week ago, Vicki.
9:12 Ah, okay. I just saw I just saw the post
9:14 today. All right. But anyway, um but but
9:18 I just want to talk about it. I mean
9:19 it's you know if you run into people
9:21 that are angry or fearful or or you know
9:25 super resistant
9:28 I think one of the interesting not one
9:30 of the interesting things I think one of
9:31 the important things to do is
9:35 you know understand you know what's the
9:37 fear where you know what's where's the
9:39 anger coming from um
9:42 a lot of times the things that people
9:45 are upset about I all like I agree with
9:48 them Right. And a lot of times they
9:52 don't know what to do with that. Um, but
9:55 my philosophy is I I have no interest in
10:00 taking sides, good or bad, about a
10:03 technology that isn't inevitably going
10:05 to progress forward. What I'm interested
10:08 in is if it's going to progress forward
10:10 and knowing what we know about it now
10:12 and knowing what we know about how it's
10:15 improved. Hang on, I got to sneeze.
10:23 Do I maybe?
10:30 Yes, I do.
10:37 If the only thing that I can be certain
10:40 about is that it's not going away and
10:42 it's going to keep improving,
10:45 then what do we do about it? And the
10:48 only thing I know how to do is learn
10:51 about it, play with it, understand it,
10:54 understand where it's strong, understand
10:56 where it's weak. I think I think one of
10:58 the things that people that are outside
11:00 of AI that are not playing with it do is
11:02 they they romant they romanticize it or
11:05 demonize it, right? The romanticizing it
11:07 of it is it's, you know, it's this
11:09 allseeing, all knowing, omnipotent
11:12 being. the demonization of it is this
11:15 it's this evil robot that's going to
11:16 kill us and take all of our jobs and and
11:18 ruin everything. Neither of those are
11:21 true. And when you use it, you go, "Oh,
11:24 I get it. It's a tool. It's really
11:26 powerful. It's got these things it does
11:27 well. It's got these things it doesn't
11:28 do well, and it's getting better and
11:30 better and better and better." And so,
11:33 what are the implications of that? One
11:34 of the implications is that as it gets
11:36 better, there's going to be a demand for
11:39 people who know how to deal with it and
11:42 know how to think about it and think
11:44 critically about it and understand what
11:46 it does good and what it does bad and
11:47 and find people that have chain of craft
11:53 with AI that is fluid and flexible and
11:56 they can pull in lots of different tools
11:57 to solve problems. Those people are
11:59 going to be in demand.
12:01 So, if you're here, how many people do
12:03 we have here? 32 and 16 on on the other
12:05 channels. Um,
12:09 you're way ahead of the game. And even
12:12 if you [ __ ] hate AI,
12:15 but you still came here,
12:18 welcome and cool
12:21 because the stuff that's shitty about
12:24 it, that is legitimately shitty about it
12:26 is shitty.
12:28 And I agree with you there. the stuff
12:31 where people actually don't understand
12:32 how it works and they've got it in this
12:34 sort of fantasized,
12:36 you know, utopian or demonic
12:39 extremes.
12:42 Learn it, understand what's there,
12:44 understand what it how it can help you,
12:47 right? If you just got laid off, you now
12:50 have access to these tools that can
12:51 amplify your ideas. I just talked to
12:54 someone today who was a senior executive
12:56 at one of the big ad ad holding
12:58 companies, really good friend of mine,
13:00 and he got laid off last March. And, you
13:02 know, he sort of bounced around trying
13:04 to find jobs for a couple of months. And
13:07 then he said, you know, the writing's on
13:09 the wall with his AI [ __ ] I'd better
13:12 get going. And he signed up for an MIT
13:13 course and he's he's in the middle of
13:15 that and he's he's getting his ass
13:17 going.
13:19 So,
13:21 so anyway, I just thought that was worth
13:24 saying saying into the universe, you can
13:27 make money with Chachi Bet. Where's my
13:30 button?
13:31 >> You can make money with Chy.
13:33 [Laughter]
13:38 [Music]
13:57 [Applause]
13:59 [Music]
14:15 Oh, tonight T tonight for the AI leaning
14:17 lab.
14:19 So, producer Brandon, uh, I've I've
14:22 assigned him spellchecking and
14:24 apparently
14:25 apparently he missed that one. That's
14:27 that's on him. It definitely wasn't me.
14:29 I don't make typos. I use AI to type, so
14:33 I never make mistakes.
14:35 [Laughter]
14:44 Oh, man. All right.
14:47 AI leaning lab. That's really funny.
14:50 Genie in the bottle. I thought another
14:53 channel was trying to steal your
14:54 thunder. No, that's just a typo. I just
14:56 screwed up.
15:01 Oh, man. All right.
15:05 How much have you generated
15:09 with the sale of that button? I have
15:11 never sold this button. I should sell
15:13 this button. You can make money with
15:14 every
15:16 >> um this was just an inside joke here. I
15:19 listen I had a I in the in the heyday of
15:23 the irregulars and all of these jokes
15:25 like where this joke came from in the in
15:28 the early days of the of AI.
15:32 >> I should have sold a bunch of merch and
15:33 I didn't do it. So at some point I'll
15:35 probably do that. But yeah, we could
15:37 sell that one.
15:39 Some we we'd probably sell six or seven
15:41 of them. make make $17.
15:47 Nice guitar strap. Thank you. I think
15:50 this is you know what where I got this.
15:52 I think this was just an Amazon guitar
15:55 strap cuz I was I had one more guitar
15:57 than I had straps and I kept rotating
15:59 them and one day I was just like, "Oh,
16:01 [ __ ] it." And I went I went to Amazon
16:03 and just ordered that. That's where I
16:04 think this is from. So, no sentimental
16:07 value at all.
16:10 [Music]
16:21 Did someone say early days? Lunick,
16:24 what's happening?
16:27 You remember the early days of
16:28 >> you can make money with Jack?
16:31 >> I'll tell you what's funny. You know
16:32 what's funny right now?
16:35 It's It's not um these these douchebags
16:39 have left, but they've been replaced
16:41 with um
16:45 What are they doing now?
16:50 It It's all the uh um you know, there's
16:54 a new image generation model out from
16:56 Gemini, and I'm going to teach you how
16:57 to prompt it. You can't just do simple
16:59 prompting. You have to do complex
17:00 prompting, JSON prompting.
17:04 You too can prompt with Jason.
17:09 [Music]
17:20 You got to start selling merch. All
17:21 right, we'll get on it. Actually, you
17:23 know, producer Brandon, we should
17:25 probably talk about doing merch. I think
17:26 we have
17:29 We've got stickers. as long as stickers.
17:34 Um,
17:38 all right.
17:40 We were in Staples yesterday and my
17:43 buddy said, "Doesn't Kyle have one of
17:44 those buttons?"
17:48 Oh, my hubby. That's really funny.
17:51 [Music]
18:03 You know,
18:08 [Music]
18:29 Oh, I want those stickers for DC. So,
18:31 Daisy, uh, DM me and we'll connect with
18:34 Tobias. I don't know if Tobias is here,
18:35 but Tobias printed a bunch of those, so
18:38 we can we can, uh,
18:41 we can get we can at least cover his
18:43 cost on those and and get some of those
18:45 for DC. When are you going to DC?
18:51 [Music]
18:52 [Applause]
18:53 [Music]
18:54 [Applause]
18:56 Oh, so by the way, if you're new here,
18:58 welcome. My name is Kyle Shannon. This
19:00 is the AI learning lab. This part at the
19:02 beginning where it's just like, does he
19:03 ever talk about anything? Sometimes I
19:06 do, but this is just kind of the
19:08 warm-up. It's like letting people, you
19:10 know, digest their food, have a little
19:12 appertif. We got a very sophisticate
19:15 sophisticated crowd here who has
19:18 appertif.
19:24 I'm pretty sure that's what it is.
19:26 Anyway, apparently people like them and
19:29 you know they like to have their meals
19:31 and then they like to have their
19:32 appertifs and then they like to come to
19:33 the AI learning lab and a lot of people
19:37 many people
19:39 um have given up Netflix to watch this.
19:46 It's true. I don't know if it's many at
19:48 least one or two. Um
19:52 but we will talk about some AI stuff
19:54 tonight. Um,
19:56 we're going to dig back into Gemini um,
19:59 image generation. Um there there's some
20:02 cool stuff that it does that I want to
20:04 experiment with where you can create you
20:06 can upload three different images. But
20:09 in each of those images you can like
20:12 give it instructions and it it can parse
20:16 because okay
20:18 the image generation models that are now
20:21 within chat GPT and within Gemini.
20:25 I'm trying to think if there's another
20:26 one that's got it in it right now.
20:29 Maybe it's just those two. Right now,
20:32 it's actually native inside the large
20:35 language model. It's not even I is it
20:37 still called a large language model if
20:38 it's multimodal? I think it is.
20:42 But anyway, so it can understand images
20:45 and audio and video like the pictures in
20:48 video and text and code all of it.
20:53 Um,
20:54 and so you can put like written
20:56 instructions on the the graphics that
20:59 you upload and it will follow those. So,
21:01 I figure we'll experiment with that a
21:02 bit tonight. I could be watching season
21:05 6 of The Secrets of Skinwalker Ranch.
21:08 You're that good, Kyle. Hey, listen.
21:10 Skinwalker Ranch is one of my favorite
21:12 shows. Like all ghost shows, my my
21:15 favorite thing about those shows is that
21:17 they never find anything.
21:20 The one the one treasure show that I
21:22 watched, it was called uh I think it was
21:25 the secret or the curse of Snake Island.
21:30 And and they started out this treasure
21:32 hunt journey on Snake Island, which is
21:35 named that because it's got the highest
21:39 concentration of snakes of any place on
21:42 Earth. It's one poisonous snake per
21:44 meter, per square meter.
21:49 [ __ ] idiots. And they went on there
21:51 to find a treasure and then and then
21:54 some hint there said you should go to
21:57 South America. I think they just wanted
21:59 to get off Snake Island. But they go to
22:01 South America and then in that season
22:03 they [ __ ] found the treasure. It was
22:06 insane. And and you you want to know
22:08 about season 3?
22:11 There wasn't season 3. They found the
22:13 treasure and they're like, "Fuck TV. We
22:15 got treasure."
22:18 So, we never got to hear about the
22:20 treasure.
22:22 Unbelievable.
22:24 So, you know, if those Ghost Hunter
22:26 shows, if they ever actually find a
22:28 ghost,
22:29 the show's over.
22:31 [Laughter]
22:34 We just watch it like a bunch of dupes.
22:36 Like, I wonder if they're going to get a
22:38 ghost this week. Did you hear that?
22:41 Something clicked. Did you see that
22:43 floater? It floated right across the
22:45 camera. I'm pretty sure it was sentient.
22:49 I love those [ __ ] shows. My favorite
22:52 one it was it's the the one with the uh
22:55 what I forget what it's called.
22:58 It's not Ghost Hunters. Maybe it is.
23:02 It's the one with the the meathead guy.
23:03 The really pumped up beefy dude. He's
23:06 like got a big voice and he's always
23:08 like getting all macho with the ghosts
23:11 and then the minute one of them touches
23:12 him he's like
23:18 But anyway, I'm glad I'm more
23:19 entertaining than those shows. Um,
23:25 I saw a couple of Tik Tok videos that
23:27 they've apparently found the treasure uh
23:29 on Oak Island, which which there if they
23:32 actually have found the treasure, um,
23:35 more power to them because they've been
23:36 doing that for more like a decade now,
23:39 11 years, 12 years. It's crazy.
23:44 Oh man. Okay. Um,
23:49 danger noodle island.
23:52 Oh, the no prop. Yeah. Yeah, exactly.
23:56 That that that curse of snake island or
23:59 whatever the hell it was called.
24:02 That was trippy, man. cuz they're
24:03 walking through cuz there's no humans on
24:05 it cuz there's that many snakes and
24:07 there it's just walk walking through
24:09 brush everywhere and like at their feet
24:12 at their waists and at their faces are
24:15 snakes,
24:17 you know, and they're like pushing
24:18 sticks away and you can see the snakes
24:20 kind of scurrying away and
24:23 it's like Jesus,
24:26 you know, TV's swell and all but really.
24:30 All right, let's go. Let's go play.
24:34 Let's go play. Anybody I I've just been
24:36 rambling. Anybody have any questions,
24:38 thoughts, anything you want me to cover,
24:40 talk about.
24:51 Should we do ASMR?
24:54 Hey,
24:56 hey, welcome to the AI learning lab.
25:06 What about cling? Um, I haven't played
25:09 with cling in forever.
25:12 Oh, what's great about clinging? Well,
25:14 so cling.
25:18 So, so Cling, I think it's Cling 2.1.
25:23 Cling 2.1, Luma Labs, Runway,
25:28 VO3,
25:29 Midjourney.
25:33 There's some Chinese ones.
25:36 There's like there's like I don't know,
25:38 seven or eight really good video models.
25:41 Cling's one of them.
25:43 Yeah. Nano Banana. Nano Banana plus
25:46 video. All right. Yeah. So, that's what
25:47 we're going to do. We're going to go
25:48 figure some [ __ ] out.
25:50 Nano banana plus video equals magic.
25:55 Cling has some promos going on right
25:57 now.
25:59 I would think that all these video
26:00 companies are going to have promos going
26:02 on because there's there's so many of
26:03 them. I I have a feeling that the the
26:06 video industry is going to get really
26:08 oversaturated quickly and and you'll see
26:11 consolidation maybe in the next two
26:13 years.
26:16 Facebook isn't letting me share my link.
26:18 That's cuz they're [ __ ]
26:21 Um, you know who's the other one that
26:23 that just promos the [ __ ] out of their
26:25 stuff is Higsfield. And I I don't know
26:28 if Higsfield has their own model or if
26:29 they're just using other people's models
26:31 and they've got some pre pre-done
26:34 special effects. It's it's it's like at
26:38 this point it's impossible to keep track
26:39 of who's got models, who's using other
26:42 people's models. I I just I can't
26:45 anymore.
26:47 I just try to pay attention to the [ __ ]
26:49 that makes the most noise. Um,
26:53 did you ever upload the guy singing?
27:00 Oh, singing. Singing the other night. I
27:02 did. Let's Let's go. Let's go look at
27:05 that.
27:07 Video still expensive. Video is still
27:09 expensive. Although right now you can
27:13 Google's given everybody the option to
27:16 make some videos with VO but it's like
27:19 they're giving you like five videos.
27:20 It's not enough. Yeah. Video video is
27:23 still very expensive.
27:26 My guess is
27:29 what's going to happen is this.
27:32 So, do you remember remember when I I
27:36 used to play with Deco here? The real
27:38 time image generation stuff. Korea's got
27:41 it. Um, Deco here has it. It's all based
27:44 on um
27:47 what the hell is it based on? Um,
27:50 SDXL.
27:53 I forget who made it. Anyway, it doesn't
27:55 matter. It's a It's a model that that
27:57 generates images really, really quickly.
28:01 They've got video versions of that as
28:02 well. So they've so so in testing in in
28:05 some of these labs, they've got video
28:07 models that are generating
28:09 video faster than 24 frames a second.
28:13 And if you can generate video faster
28:16 than 24 frames a second, you can start
28:18 to do real time stuff. and and the uh
28:22 the Genie 3 that that Google just
28:24 announced, which is the Speak a World
28:27 into Existence, you're walking through a
28:30 3D world that's actually being generated
28:32 at faster than or it's being generated
28:35 at 24 frames a second. Um,
28:39 it's not showing you
28:42 my input and output. What?
28:49 All right, you're talking to each other.
28:51 Black bar. Wait, and don't I need to
28:54 show Someone wanted to know about the
28:56 song. So, let's let's go.
28:59 Let's go find the song. So, this was the
29:01 thing that I tried to upload the other
29:03 night. I I just couldn't. And then right
29:06 before Brandon and I hung up. Oh, this
29:09 was funny. This was kind of cool. So, I
29:11 took this image. So, this is like my
29:13 corporate image, my corporate head shot
29:15 from like, you know, seven or eight,
29:17 nine, 10 years ago. I forget when we
29:19 took it. So, I uploaded that into um
29:24 into Nano Banana into Gemini 2.5 Flash.
29:27 And normally whenever I've used that
29:29 image to train someone before, my head's
29:32 always in that position and it's always
29:34 got that dorky smile on it. And so what
29:37 I wanted to do was see if like I could
29:39 get it to retain sort of what I looked
29:41 like, age me up a little bit, um and
29:44 change my expression and change my
29:46 position. And so I think it took me
29:48 seven or eight prompts, but I ended up
29:50 going from that to that,
29:53 which is, you know, that's pretty cool.
29:56 I look hip.
29:58 I look hip and like I need a haircut.
30:06 Oh man.
30:09 Um,
30:11 but the other night, let's see, what did
30:14 I make?
30:16 I made that dude in Nano Banana.
30:22 There we go. That's the one. Tab
30:25 hoarders unite.
30:30 I'm sharing the right tab. I think I am
30:33 of a thousand favicons.
30:37 [Music]
30:46 I bow to the altar of a thousand favic
30:48 kinds. Each a tiny stained glass
30:51 promise. Each tab relic an argument. An
30:55 unfinished prayer. We do not close, we
30:57 collect.
30:59 We are archaeologists of distraction.
31:01 Curators of half-bun obsessions. We
31:03 preserve the tabs that you discard.
31:07 They call us tab hoarders like it's a
31:09 crime. I call us inspired keepers of the
31:12 incandescent. Now
31:14 every tab is a shrine to a thought that
31:17 refused to die. A headphone cable
31:19 tangled with yesterday's revolution. A
31:22 recipe we swore we'd master. A thread we
31:24 swore we'd answer. We keep the world in
31:27 micro windows like sailors with sex
31:30 mapping storms of curiosity.
31:34 Chrome hates us. Chrome groans under our
31:36 little empires. It's memory like an
31:38 anxious brain on the subway. Leaking,
31:40 forgetting, dropping passengers on the
31:43 line.
31:44 Memory leak, they say, as if sin were a
31:48 bug fixable with a patch.
31:50 Google with its polished teeth and
31:53 corporate himnels cannot fathom the
31:55 sanctity of our open questions. It
31:57 trims, it prunes, it gaslights our tabs
32:00 into sleeping bags. But we are stubborn,
32:04 gentle, glorious, unruly.
32:09 [Music]
32:11 Let the browsers call it inefficiency.
32:13 We call it fidelity. We are a million
32:16 tiny vows. I will return. I will return.
32:19 I will return. Each tab is a bookmark
32:22 for an argument with the future.
32:24 We keep versions of ourselves and
32:26 duplicate. The person we were at 2 a.m.
32:29 The person who clicked that headline in
32:31 a cafe. The person who loved a sentence.
32:34 Close one and you risk killing that
32:36 small particular grief. Close one and
32:38 the map of you becomes less accurate.
32:41 There is an ethics to tap hoarding.
32:44 Not greedy. Not.
32:45 >> So anyway, that's that uh question. What
32:47 website is this? This is just on Twitter
32:49 right now. So actually here's here's a
32:51 request. Go to my Twitter account, Kyle
32:54 Shannon. S Han N O N. and find this
32:58 post. It's called Tab Hoarders Unite.
33:01 It's like, I don't know, six or seven
33:03 posts down. And do me a favor and like
33:05 go like it, quote post it, do [ __ ] like
33:08 that because it's kind of fun. I like
33:10 it. And I I like the the rebellious tab
33:13 hoarder. Um, if you're wondering how
33:14 that was created, um, I wrote the
33:19 lyrics.
33:23 I wrote the lyrics I think in chat GPT,
33:25 but it might have been Gemini. It was
33:27 one of the large language models, Gemini
33:29 or Chat GPT.
33:32 And then um
33:35 and then went to I was experimenting
33:38 with um Sunno to see if Sunno could do
33:41 spoken word and it can. That's that's
33:43 what it was or that's what it did.
33:47 Um, and then I took the audio file. Um,
33:52 and I took an old picture that I had
33:55 from my Kyle Shannon Dreams project,
33:59 which is that, you know, sort of beat
34:01 poet
34:02 washed up 60-year-old.
34:07 And uh I took that image into um
34:13 into Nano Banana into Gemini 2.5 and um
34:18 made this thing or maybe I did it in
34:20 midjourney. I think I did it in
34:22 midjourney. Yeah, I did it in midjourney
34:24 and then I took this image and the sun
34:28 and put them together in Hedra.
34:31 All right.
34:36 Hedra, Hedra, Hedra, Hedra does the
34:38 animations pretty well. Um, Hen today
34:42 came out with a new version of their
34:44 digital twin technology. So, for what
34:47 that's worth. Oh, thanks. It looks like
34:48 a bunch of people have gone and shared
34:50 that. So, yeah, if you go if you go to
34:52 my Tik Tok uh channel or no, my Twitter,
34:55 my ex channel, Kyle Shannon, and find
34:59 the Tab Hoarders Unite song and just
35:01 boost that. That would be swell because
35:05 X
35:07 when whenever I post anything that I
35:09 think is really cool and like people are
35:11 going to see this. It gets like three
35:13 clicks like God
35:16 social media sucks.
35:20 Um I guess we'll just change this tab
35:24 into a Gemini tab.
35:29 Okay. So where I am right now is at
35:31 gemini.google. google.com
35:34 and I'm on the 2.5
35:36 flash model.
35:39 So 2.5 flash has got this new image
35:42 generator tool in it
35:45 or as they call it on the X
35:50 it's nano banana.
35:53 Um,
35:56 so this is here and then we can upload
35:59 up to three image files. So I think what
36:02 we're going to do, I'm going to change
36:03 how I'm sharing here for a second. I'm
36:05 going to switch over to sharing my whole
36:08 screen.
36:18 down.
36:31 All right.
36:33 All right. Fantastic. Bob,
36:36 have they won anything? Uh, no. It's
36:38 It's not a game show. Um, okay. So, I'm
36:43 going to go there. Then I'm going to go
36:45 I'm going to go to here.
36:47 What's this? This is
36:51 [Music]
36:53 This is Keynote.
36:57 So, I'm going to make a new slide.
37:01 All right.
37:09 All right. So, here's what we're going
37:11 to do.
37:22 Um,
37:24 okay. Let's go find
37:27 photos. Um,
37:30 so what I'm doing,
37:33 what I'm doing right now,
37:36 let me let me explain the madness that
37:39 is about to ensue because it's not
37:42 really going to be madness. it's just
37:43 going to be boring and and if but if you
37:46 know if you know what I'm doing it might
37:48 be a little less boring. Um
37:53 so right now I know I want to explore
37:56 this image generation model in Gemini. I
37:59 know that you can upload three different
38:01 images and tell it to combine them. But
38:03 I also know that those images can be
38:05 quite sophisticated. So, so what I'm
38:08 trying to do right now is figure out
38:12 what I want to do. I don't really have a
38:14 clear use case.
38:16 So, for the next 15 minutes or so, I'm
38:18 just going to be bumbling
38:20 and putting together the piece parts
38:22 that we're then going to take into Nano
38:24 Banana and see if it can parse them. Um
38:28 because again, part of the part of my
38:31 philosophy, so if you're if you're new
38:34 here, hey Brandon, can you pop up the um
38:39 can you pop up uh the AI salon banner?
38:44 Yeah, there you go. So, if you're new
38:46 here, um
38:52 [Music]
38:55 All right, that's still there. Okay,
38:56 great. If you're new here, go to the
38:58 salon.ai. Actually, if you go to
39:00 community.thesalon.ai,
39:03 that'll take you to the community site
39:06 of the AI salon. So, this is the the AI
39:09 learning lab is an extension of the AI
39:12 salon. It's it's a part of it. So,
39:14 basically, this channel is brought to
39:15 you by the AI salon. Um,
39:20 it's a really amazing community of
39:24 AI friendly, curious, adventurous people
39:27 that are trying to figure this stuff
39:29 out. Um, so a lot of the a lot of the
39:33 work that we'll do in here that I'll
39:35 play with in here, I encourage you to
39:36 play along. And one of the things that
39:38 we talk about in the AI salon is what we
39:41 call the cycle of AI readiness. And let
39:43 me just I'm going to just pop over there
39:46 for a second and talk about it because I
39:47 think it's really important.
39:51 And the cycle of AI readiness looks like
39:54 this.
39:56 Play first, mindfully create, generously
39:59 lead.
40:00 And so what I'm doing tonight is is play
40:03 first. I'm just playing. So I don't have
40:05 an agenda. Mindfully create is like once
40:08 you learn how to do something and once
40:09 you've got some basic skills in a tool
40:13 um then you can mindfully create. You
40:15 can go problem solve or you can go do
40:17 you know self-expression.
40:20 Um but but the stage we're at tonight is
40:22 just this play first. And so what I'm
40:23 going to be doing is just kind of
40:24 [ __ ] around. So it it might if it
40:27 seems like I don't have a clue what I'm
40:29 doing, that's because I don't have a
40:32 clue what I'm doing.
40:35 And if part of my role here on the AI
40:37 learning lab is to make it okay for you
40:40 to not know what you're doing, then I've
40:42 done my job.
40:44 I'll be clueless first. You follow
40:46 along. Um, and you know, don't don't
40:50 shame yourself for not having your head
40:52 around this stuff because like nano
40:55 banana, this new image model came out
40:56 two days ago and part of what we're
40:58 trying to figure out is is it any good?
41:01 And the answer is yes. But it's like,
41:03 okay, how specifically is it good? So
41:05 maybe tonight we'll play around with
41:07 some JSON prompts. Maybe tonight we'll
41:08 play around with this image prompting.
41:10 Things like that. All right, make good
41:12 sense. So go join the AI salon. Step two
41:15 in the AI salon is introduce yourself.
41:18 So if you haven't done it yet, if you're
41:20 here just lurking and you haven't
41:21 introduced yourself in the AI salon,
41:24 shame on you.
41:26 Don't be a ninny.
41:29 Okay. Um,
41:32 where was I going to go? I know where I
41:33 was going to photos.
41:40 So, we're going to go find
41:54 So, these images here are the part of
41:56 this thing I created called Kyle Shannon
41:59 Dreams.
42:01 So, what we're going to do is we're
42:03 going to just Can I copy that? Why can I
42:07 not copy that?
42:09 Copy. Can I go over here and paste it?
42:13 Yes. Okay, great. So, we're going to get
42:16 some weird
42:20 weird things. Copy.
42:23 Oh, wait. We'll do it like this.
42:26 Get that guy. We'll get that woman.
42:31 and we'll get him. Okay. Copy. Okay.
42:34 Now, we're going to go over here. Okay.
42:36 So, here's what we're going to do.
42:44 Okay. Okay.
42:47 We're gonna we're going to name these
42:49 these these people.
42:54 [Laughter]
42:58 I'm such an idiot.
43:03 This was a fun project. This was a
43:05 project I did. All these images came
43:07 from uh from 2022
43:10 in the fall. Um, I taught myself to
43:13 install stable diffusion
43:16 and
43:19 and then I
43:21 I installed this thing called Dream
43:23 Booth, which allowed you to upload some
43:26 images of yourself and and make these
43:28 crazy ass self-portraits.
43:35 Um, how are we going to do this?
43:45 Oh, I know. Don't do
43:55 shrink.
44:00 That singing he does is very annoying.
44:06 All right, so let me grab all of these
44:08 things. Bring them to the front.
44:12 Let's lock that bad boy.
44:15 Okay, now I've got more room to play
44:16 with. Now this makes it easier. Okay, so
44:19 we got those three and then we got these
44:22 three. Right.
44:27 Okay.
44:31 Put that there. We'll put that down
44:32 there. All right. Now we'll add some
44:34 text. And now we're going to name these
44:36 things.
44:38 So, this one we're going to name Willie,
44:40 as in Willy Wonka.
44:42 [Laughter]
44:46 We're going to name this guy Tom.
44:52 We're going to name this guy Benson.
44:55 Benson.
45:00 Uh,
45:02 ideas. Jared. Jared. That's Jared.
45:08 And and you're like, Kyle, why are you
45:10 naming them? I just cuz I'm playing. I'm
45:13 playing. People, don't you understand
45:16 play? Uh, this is the mad hatter. What
45:19 should his name be? Um,
45:25 um, Haddington.
45:26 [Laughter]
45:29 That's original. Um, we're going to call
45:32 her, uh, Bessie.
45:36 And then we're going to call this guy
45:39 Tony as in obnoxious award recipient.
45:45 Tony. Tony award winner.
45:49 Okay. Willie, Tom, Jared, Haddington,
45:53 and Bessie. Okay. So, let's duplicate
45:56 this.
46:01 Okay. I got an idea. Okay. Got it. Got
46:03 it. Got it. Going back to my picture
46:05 things. Now I'm going to go
46:11 Kfax and Wazi exhibit.
46:15 Okay,
46:17 we're going to grab
46:22 this sunset.
46:25 Copy.
46:29 Get rid of all these things.
46:34 going to put in that sunset
46:38 that was this was on the uh the
46:40 mountaintop right above Boulder,
46:42 Colorado. This was I was driving to
46:44 Boulder one night and that sky
46:48 was there was amazing. All right.
46:52 So, we're gonna I don't think we need to
46:54 label this one, but we're gonna have
46:57 Okay,
46:59 we're gonna have that.
47:02 And then we're going to have
47:05 We're going to find a place.
47:08 Ah, this this is a good one. Okay. Copy.
47:13 Go here. This was Oh, this is also
47:15 really cool. So, so when I took this
47:17 picture, this motel picture,
47:22 so you really can't see it, but right in
47:26 between these two cars, there's this
47:27 little fold up chair. And so, I was just
47:30 in the parking lot. I had my camera out.
47:32 I was taking pictures cuz the sky was
47:33 really cool. And
47:36 um and I saw this sign and I wanted a
47:38 picture of the sign. And as I pulled in,
47:39 there was there was this trashy woman
47:43 sitting in that chair smoking a
47:45 cigarette. And I tried to discreetly
47:48 pull up my camera and get a shot of her,
47:49 and she just immediately took a beline
47:52 into her room. It's still a cool shot,
47:55 but it would have been cooler with the
47:57 trashy woman smoking a cigarette there.
47:59 But you know, what are you gonna do? All
48:00 right.
48:04 All right. So, okay. So, here's what
48:06 we're gonna do. We're going to we're
48:07 going to
48:11 we're going to take screenshots of these
48:13 things,
48:15 save them to the desktop.
48:26 Okay.
48:29 And then I got to remember these names.
48:31 Hey, do me a favor, Brandon. Can you
48:33 just write down these names?
48:36 Um, Willie, Tom, Jared,
48:42 Haddington, Bessie, and Tony.
48:50 I love, by the way, in case you didn't
48:53 know it, one of the things that gives me
48:55 great joy in life is just being an
48:58 idiot.
49:01 Okay. All right. So, that's saved. Okay.
49:04 So now we have some input images. You're
49:07 like, "But Kyle, where is this going?
49:09 This is so fascinating. H what are you
49:12 doing? I don't know."
49:15 Okay,
49:17 so we're going to add files.
49:20 So I got to go find screenshot.
49:25 There should be two new ones. Boom.
49:27 Right. Yep. There's that. There's that.
49:30 Okay.
49:32 So, what we're going to do
49:36 is we're going to say
49:40 I want
49:42 um
49:45 a group photo
49:49 of
49:54 Willie
50:00 Tom
50:02 Wait. Willie, Tom,
50:05 and Haddington
50:13 right
50:15 in front of the camera.
50:24 Arm in arm.
50:27 I want
50:31 Willie, Tom, and Haddington.
50:33 So Jared, Bessie, and Tony. I want
50:37 Jared,
50:41 Bessie, and Tony
50:44 to be
50:48 spread out across
50:51 the
50:55 second story balcony of the hotel.
51:05 motel
51:07 looking
51:10 down on the threesome taking their
51:14 picture
51:17 and I want
51:19 the sky
51:22 to be that crazy
51:26 sunset.
51:28 Okay,
51:30 so in theory, we're going to have three
51:33 people standing in front of the camera,
51:35 the hotel in the background, three
51:38 people up on the second floor balcony,
51:40 and the crazy
51:42 and the crazy sunset.
51:45 Oh, and then we'll do this.
51:48 Make the image
51:53 16x9 wide aspect
51:57 ratio.
51:59 Make sure.
52:04 No, I'm not going to say make sure
52:06 anything. [ __ ] it. Let's see what it
52:07 does. Right. Okay. We're at
52:10 gemini.google.com google.com
52:14 playing with 2.5 flash or nano banana
52:22 and we're going to see if it can
52:23 actually understand and parse those
52:24 images. All right, here we go.
52:28 [Music]
52:32 Uh uh uh uh uh uh uh. Now, what's really
52:39 powerful about this if it works
52:42 is here we go. Whoa.
52:48 Okay, we didn't quite It sort of [ __ ]
52:51 up a little bit.
52:58 It put them on the third floor. Okay.
53:00 So, so, so, so it
53:07 it kind of it kind of worked. I mean, I
53:10 didn't give it the best instructions,
53:12 did I? Um,
53:19 all right. That's kind of cool. So, so
53:21 it made Tom How do I zoom in on these?
53:25 So it made it made Tom. So Tom, this
53:28 guy,
53:30 the guy in the middle has Willy's outfit
53:33 on, but that's definitely Willie. That's
53:35 definitely Haddington.
53:37 And that's definitely Tom. That's cool.
53:41 And then up here you have you have
53:47 it put Haddington in place of the woman.
53:51 Did I put Did I put Bessie? Did I put
53:53 that right? Hang on.
54:04 Yeah. Jared, Bessie, and Tony. Okay. So,
54:07 now we're going to do this. Now, we're
54:09 going to say, um,
54:13 I want
54:16 a closeup
54:19 of
54:23 Bessie
54:25 on the roof
54:28 near the motel
54:31 sign with the
54:38 Um,
54:40 sunset
54:42 in the background. This is pretty cool,
54:45 isn't it?
54:50 All right, let's see what it does with
54:52 that.
54:55 We'll see if it remembers Bessie.
55:05 All right. No, it didn't. It didn't
55:07 understand that. Okay.
55:11 It just gave me that image again. That's
55:14 weird.
55:16 Okay. So, we'll try this. We'll try
55:21 Oh, you know what I'll do? I know what
55:23 I'll do. No, Kyle. What?
55:26 We'll re-upload those two images.
55:29 I'm going to say, "Okay,
55:32 totally new image idea.
55:37 I want Bessie
55:40 by herself on the roof.
55:48 The camera,
55:50 the the
55:52 photographer
55:56 is on the roof
55:59 with her and
56:02 the
56:04 motel signs
56:08 are behind and above her.
56:14 The sunset sky is still
56:19 dramatic.
56:22 All right, let's see if it'll just take
56:24 one of those characters
56:27 and give us a new thing without having
56:29 to start a new chat.
56:45 Hey Ryan, let's cool it down, please.
56:46 Thank you.
56:54 So, did it do Bessie?
57:03 Yeah, it did.
57:06 Well, sort of.
57:11 Yeah, that sort of looks like her.
57:14 Wow, that's pretty remarkable.
57:21 So, this [ __ ] works.
57:26 All right, so let's let's go get a
57:29 different sky.
57:33 [Music]
57:43 Uh
57:59 black bar. Black bar. Oh, blackberry
58:03 bamb. Oh, blackberry bama jam. All
58:06 right, there we go. There we go.
58:13 Mine didn't work. It gave me the code
58:15 for the image.
58:17 Yeah. What I'm noticing about nano
58:19 banana is like sometimes if I say
58:24 um you know have these three standing
58:27 near each other, it doesn't understand
58:30 that it wants you to make a picture. So,
58:32 I'm like I'm finding myself having to go
58:34 make a photograph of like you have to
58:36 really expl be really explicit with it.
58:39 But, let me I'm going to go grab a
58:40 different um
58:44 location.
58:49 I think this one's cool.
58:52 This barber shop. Oh, we'll do Union
58:54 Station. Okay. Copy. Okay.
59:00 So, we got Union Station.
59:05 I think we got room for one more
59:06 location, don't we? We sure do.
59:14 Sorry. Sorry. Carrying on rather Cheerio
59:17 Pip. Okay, we'll get the
59:21 barberh shop.
59:30 When I retry, it says something went
59:31 wrong. It showed me the thinking then
59:33 the code. That's weird. Um,
59:37 I would Oh, hang on. Physical dexterity
59:40 on tik tok. Oh, I missed it. Sorry. Hang
59:42 on.
59:44 All right. Um,
59:52 okay. So, we got three locations, two
59:55 skies. Let's get a third sky. Three
59:57 locations, three skies.
1:00:00 Oh, we'll see if it can do logic. That
1:00:02 would be cool.
1:00:04 That would be cool. Um,
1:00:08 let's get a golden sunset. Okay, here we
1:00:12 go. this one.
1:00:14 Copy paste.
1:00:19 By the way, if you want to see these
1:00:20 photographs, most of these photographs
1:00:22 are at um
1:00:25 if you go to kyleshannon.com,
1:00:27 that's my photography website. These are
1:00:30 actually made with a camera and my
1:00:32 eyeballs.
1:00:36 And and I gotta tell you, you know, it's
1:00:38 funny when I got when I got hate blasted
1:00:41 on LinkedIn for saying that you could
1:00:44 possibly be creative with AI.
1:00:47 Um,
1:00:50 you know, people were just saying, you
1:00:51 know, you can't possibly be creative
1:00:53 because it's just pushing pushing a
1:00:54 button. So is [ __ ] photography.
1:00:58 Like
1:01:01 the reason
1:01:03 I was talking to a friend once because I
1:01:05 couldn't figure out what what the
1:01:07 through line was of my photography
1:01:09 because it all felt very different.
1:01:11 And he said to me, he goes, "Oh, no. You
1:01:13 you've absolutely got a a photographic
1:01:16 voice." And I said, "What is it?" And he
1:01:18 said, "You take pictures of the things
1:01:21 that most of us walk by and don't
1:01:23 notice."
1:01:25 And I was like, "Oh, that's so cool."
1:01:27 And and I find when I'm when I'm doing
1:01:29 image generation in AI, when I'm just
1:01:31 sort of [ __ ] around with it,
1:01:34 it has a very similar feeling as to when
1:01:37 I'm out with a camera that I'm just like
1:01:42 sk always scanning and I'm like, is that
1:01:44 a picture? Is that a picture? I'm like,
1:01:45 oh, that's a picture. And then I'll stop
1:01:46 and I'll take it and I just take a bunch
1:01:48 of them. And then when I get back, I
1:01:50 look and see if anything was actually a
1:01:52 picture. And I feel like that's how it
1:01:54 is with AI, too. That you just generate
1:01:56 all this [ __ ] It's like walking around
1:01:58 the world and just like there's a piece
1:02:00 of gum on a bench. There's a dude that
1:02:02 looks like he's a lion. There's right
1:02:05 there's two pe people arguing. There's
1:02:07 that looks like a Renaissance painting.
1:02:10 You know, look at that sunset. And when
1:02:13 you're doing AI image generation, it's
1:02:14 the same thing. It's just like massive
1:02:16 amounts of input and then you can pick
1:02:19 and choose just like you do with
1:02:21 photography. So anyway, for what that's
1:02:22 worth. Okay, let's
1:02:28 screenshot this bad boy. So now we got
1:02:31 three locations, three skies
1:02:34 back to here. Okay, so we're going to go
1:02:38 upload files.
1:02:42 There's that. There's our people. Okay,
1:02:48 I want a
1:02:53 black and white
1:02:56 portrait
1:02:58 of
1:03:02 Tom
1:03:05 standing in front of
1:03:10 the
1:03:12 is it barber shop? Hang on a second.
1:03:17 Yeah. Hair and nail.
1:03:22 G&G up down up down. G&G uptown cuts.
1:03:27 Yeah.
1:03:30 Okay. in front of the barberh shop
1:03:37 with the
1:03:42 blue sky
1:03:49 above it
1:03:52 and then in parenthesis everything
1:03:55 should
1:03:57 let's see
1:04:00 Um, I just want the shapes
1:04:06 of the clouds
1:04:08 in black and white. Black bar. Got it.
1:04:12 Black bar. Black bar. Black bar. Black
1:04:15 bar. All right. Let's see if we get a
1:04:19 portrait in black and white of Tom,
1:04:22 who's the little comedian dude. So,
1:04:25 wait. This
1:04:29 This is Tom.
1:04:35 [Music]
1:04:41 [Music]
1:04:49 Oh, interesting.
1:04:51 [Laughter]
1:04:52 It sort of did it.
1:04:55 It definitely gave us Tom.
1:04:58 It definitely gave us Tom in black and
1:05:00 white with the blue sky behind it and
1:05:03 black bar.
1:05:06 Um,
1:05:10 that's pretty amazing. But it left all
1:05:11 this other stuff.
1:05:14 It It redid some of them. That's crazy.
1:05:17 Oh, look. And it took the It took the
1:05:19 people out of this one.
1:05:25 Okay.
1:05:28 Okay. I got I got an idea. Um
1:05:33 Okay. New picture.
1:05:40 Have all six people hanging out
1:05:48 near the guy on the bench.
1:05:53 at Union Station.
1:05:57 I don't want any
1:06:01 other
1:06:02 locations or skies.
1:06:09 Just a new
1:06:11 version of that of
1:06:17 that picture with
1:06:21 The six people added,
1:06:48 I can't pass something cool and not take
1:06:49 a picture. I know. Me, too. It's It
1:06:52 drives my family [ __ ] crazy because
1:06:54 I'm like, "Wait, wait, wait."
1:06:59 All right. It completely screwed this
1:07:01 up. But it's kind of fun how it screwed
1:07:04 it up.
1:07:13 Who's this guy?
1:07:16 That guy's That guy doesn't exist.
1:07:21 Put Willie up there. This is wild.
1:07:25 It's just [ __ ] wild, man. Crazy.
1:07:31 All right, let's try a new thing. Let's
1:07:33 try a new thing. Okay,
1:07:36 we're going to take
1:07:39 these two.
1:07:42 I'm going to duplicate this.
1:08:14 All right.
1:08:16 We're not going to label anything.
1:08:25 We're going to do this as a new
1:08:28 a new chat.
1:08:33 We're going to upload
1:08:37 the thing we just made. So, we're going
1:08:38 to say I want a
1:08:43 Oh, sorry. Tik Tok pin.
1:08:47 Kyle, thank you for always being willing
1:08:49 to take one for the team
1:08:52 and put yourself what's that say out
1:08:55 there. No, listen. I appreciate the kind
1:08:58 words, Source Camp. I
1:09:04 This stuff is [ __ ] intimidating.
1:09:08 If you don't think AI is intimidating at
1:09:10 this point, listen, here's So, there's a
1:09:12 couple of things going on.
1:09:16 When chat JPT first came out and when
1:09:18 there was just basically a handful of
1:09:20 image generation tools and some
1:09:22 animation tools and there was kind of
1:09:24 chat GPT and nothing else, it was really
1:09:27 easy to pay attention to what was going
1:09:29 on. But even back then, people were
1:09:31 intimidated by AI because they thought
1:09:34 you had to be an engineer to be able to
1:09:36 do it. Still no black bar. I know it's
1:09:38 cuz I'm a bad man. Um,
1:09:44 and that wasn't true. And a big part of
1:09:46 what this channel was in the early days
1:09:48 was this shit's easy. Just use your
1:09:50 words. You don't need you don't need
1:09:52 math and science and engineering to do
1:09:54 this.
1:09:57 What's happened in the in the two and a
1:09:58 half years since is that there's so much
1:10:00 progress and there's so much advancement
1:10:03 and the these tools have evolved so much
1:10:07 that you don't have to be a programmer
1:10:09 or an engineer still, but you do have to
1:10:12 [ __ ]
1:10:14 put some hours in. I I heard um I saw a
1:10:17 a a post on LinkedIn today that that I
1:10:20 thought was really good because because
1:10:22 I talked um I don't know a year or so
1:10:24 ago about the 10,000 hour rule is dead.
1:10:29 Meaning that as AI tools become more and
1:10:32 more capable, you don't need to put in
1:10:34 10,000 hours, you know, drawing or
1:10:36 taking pictures to get good at
1:10:38 photography or drawing, right? You can
1:10:40 just do it with a prompt. and and
1:10:42 someone on LinkedIn today said, um,
1:10:46 10,000 prompts is the new 10,000 hours.
1:10:50 And I was like, ah, that's really good.
1:10:53 And and so that's that's what we're
1:10:55 playing with here. Like, there are
1:10:58 probably an infinite number of ways that
1:11:00 I could be prompting these images. Um,
1:11:02 what we're just experimenting here using
1:11:04 image prompts and trying to get it to do
1:11:07 something.
1:11:08 This is just an experiment. the there
1:11:10 are there are people are going to find
1:11:12 all sorts of ways to exploit this. So,
1:11:16 I hope you're playing along tonight.
1:11:18 Just just play along. And if you're not
1:11:20 using these tools, just play with other
1:11:21 tools. Doesn't really matter. Put in
1:11:23 your 10,000 prompts.
1:11:27 When you say Tom, do you mean Tom Lake?
1:11:30 Looks a little like him. Uh, I don't
1:11:32 know. I just he looked like a comedian,
1:11:35 so Tom seemed like a good comedian name.
1:11:37 Okay. I want a
1:11:42 classy
1:11:45 vertical photo
1:11:50 graph
1:11:55 of
1:11:57 the man and the woman
1:12:03 standing in front of
1:12:08 Union Square station
1:12:16 with the sky.
1:12:19 The golden
1:12:21 color
1:12:23 which
1:12:26 completely
1:12:28 changes the color of
1:12:31 Union Station.
1:12:37 Not Union Square Station. That's New
1:12:39 York. Union Station. All right. Um,
1:12:44 this should be a single coherent
1:12:51 photograph.
1:13:00 [Music]
1:13:08 Where I know I'm going to do
1:13:12 a sec
1:13:35 Okay, let's see. Did we get it? Oh, we
1:13:38 did.
1:13:42 How cool is that?
1:13:48 Union Station
1:13:50 Traven and Train instead of Travel and
1:13:53 Train.
1:13:55 [Laughter]
1:14:01 There you go. All right, people.
1:14:06 We're This is This is kind of a thing of
1:14:09 another, you know, another realm. So So
1:14:11 now you can start to do stuff like this.
1:14:14 So now you can just start to to art
1:14:16 direct this. You can say, "Um, that's a
1:14:20 little too golden.
1:14:24 Let's
1:14:28 Add
1:14:30 some reds.
1:14:37 Black bar. Yeah. Blackberry. Buy my jam.
1:14:46 Nice.
1:14:48 Beautiful. Um,
1:14:52 put her in a blue dress
1:14:59 and hang on. Let's go look at um
1:15:05 we're going to go Google images and
1:15:07 we're going to go uh
1:15:11 2015
1:15:14 Jaguar
1:15:17 XF maybe.
1:15:22 All right, that's what it looks like.
1:15:24 2015 Jaguar XF. Okay, back to here. Back
1:15:29 to here. No, back to here. Yes. Okay.
1:15:34 Put her in a blue dress
1:15:37 that matches
1:15:40 the
1:15:43 2015 Jaguar.
1:15:48 What was it called?
1:15:52 XF.
1:16:04 They
1:16:08 are
1:16:13 standing in front of
1:16:18 and lighten
1:16:20 up the
1:16:23 exposure.
1:16:29 I know that's a cool sky, right?
1:16:32 Amazing.
1:16:40 There you go. There's the Jaguar XF.
1:16:43 She's now in a blue dress.
1:16:46 Dang, man. This is something.
1:16:50 This really is something.
1:17:01 Um,
1:17:03 put the car on a proper
1:17:09 3/4 angle
1:17:11 farther behind them.
1:17:27 Trying out Google's new photo images,
1:17:29 huh? Yeah, exactly.
1:17:31 Oh, it didn't it didn't understand the
1:17:33 threequarter shot of the car. Um, I want
1:17:37 to see
1:17:40 the front of the car.
1:17:44 on an angle.
1:17:48 So
1:17:52 it adds
1:17:55 depth behind them.
1:18:00 Make her hair
1:18:03 straight
1:18:06 and his
1:18:09 hair gray.
1:18:12 Uh,
1:18:16 I think that I like the dress matching
1:18:17 the sky, though. His suit could match
1:18:20 the car, maybe. All right, let's we'll
1:18:23 go back to that after this one.
1:18:27 Everyone's an art director. No, but I
1:18:29 think you're right. It's It's still
1:18:30 okay. So th this is a thing that I
1:18:33 noticed that
1:18:36 Gemini seems to get like this
1:18:40 composition stuck in its head and it
1:18:43 just can't not do that.
1:18:48 Let me try this. I'll try um let's make
1:18:52 his suit
1:18:55 dark blue.
1:18:57 Return
1:19:00 her dress to gold
1:19:05 and put the car
1:19:10 farther in the distance
1:19:13 at an angle.
1:19:16 See if that if if it does that.
1:19:22 [Music]
1:19:26 If you give the photo back to it, it
1:19:29 seems to jump start it. Yeah, it might.
1:19:32 Can you just say it car a 45 degree
1:19:34 angle? Yeah, it could.
1:19:37 Yeah. So, he's not in a blue tux. She's
1:19:41 sort of in a gold dress. And the car the
1:19:44 car is now stuck.
1:19:49 All right, we're gonna do Okay, hang
1:19:52 hang on, people.
1:19:55 I'm putting in my 10,000 prompts. New
1:19:58 chat.
1:20:00 Upload files.
1:20:07 Um,
1:20:13 the two of them
1:20:17 are
1:20:19 standing for an elegant
1:20:23 portrait.
1:20:26 Behind them at a 45 degree angle is a
1:20:34 1969
1:20:36 GTO Judge.
1:20:45 Um,
1:20:47 the sky is golden
1:20:52 and
1:20:55 they are in front of Union Station.
1:21:02 All right, let's see what we get here.
1:21:07 Got to go with a 69 GTO Judge. What are
1:21:10 we doing with a 2015 piece of snot
1:21:13 Jaguar?
1:21:19 Yeah, that's what I'm talking about,
1:21:22 baby.
1:21:28 Yeah. See what I'm saying?
1:21:33 Got that tricked out.
1:21:40 Yep, this works. Union Station. Even the
1:21:43 N is is uh is out just like the real
1:21:46 that real photograph.
1:21:50 Pretty amazing. You want to see
1:21:51 something? You want to know what's
1:21:52 what's most annoying to me about this?
1:21:56 You [ __ ] kidding me? You're going to
1:21:58 watermark all these images?
1:22:01 For as much as I pay you, Google, you're
1:22:03 going to stick a [ __ ] watermark on my
1:22:05 image?
1:22:08 What are we 1997?
1:22:11 We're going to put watermarks on
1:22:12 everything so everybody knows it's us.
1:22:14 It's called branding. We have a branding
1:22:16 specialist. And what the branding
1:22:18 specialist did was he said, "Why don't
1:22:19 you take your little AI symbol, the
1:22:21 little star, and you just put that in
1:22:23 the corner? No one will even notice it."
1:22:26 No, no, we notice it. It's a dick move.
1:22:30 Get rid of it.
1:22:35 Yeah, I know Photoshop takes care of
1:22:36 that, but I shouldn't have to [ __ ] do
1:22:38 that.
1:22:40 That is not the Dukes of Hazard car.
1:22:42 This is not the Dukes of Hazard car. But
1:22:45 we can say um replace the car with the
1:22:51 Dukes of Hazard car.
1:22:56 [Laughter]
1:23:02 Hey, Google needs brand recognition.
1:23:04 They might be forgotten. I'll tell you
1:23:06 what, Google's catching up. I mean, this
1:23:09 image gen like this this makes um
1:23:13 this makes Open AI's image gen right now
1:23:16 look like [ __ ] because um Open AI can't
1:23:18 retain characters across generations.
1:23:22 Um it's got that piss filter where
1:23:24 everything's that that bizarre antique
1:23:26 yellow. Um it's slow. It's OpenAI's
1:23:31 image gen is really slow. Um, it is good
1:23:34 at text text I would say better than
1:23:38 this this Google image gen. But but
1:23:41 these guys they're they're getting
1:23:42 ahead.
1:23:45 Oh, the green dress with Daisy Dukes.
1:23:47 Oh, that's good. That's really good. Uh,
1:23:49 wait. Oh, that goes against my
1:23:52 guidelines. Um, oh crap. Uh, make it the
1:23:57 uh what was it? The 69.
1:24:01 No, it was the it was the
1:24:04 1972
1:24:07 Trans
1:24:09 Am with uh uh black with eagle on the
1:24:15 hood.
1:24:17 The smoky and the bandit one.
1:24:26 Just a sec. Generating image. All right,
1:24:29 we got past that.
1:24:34 There we go. That's a Smokeoky and the
1:24:36 Bandit car.
1:24:42 I mean, not for nothing, but
1:24:46 good lord, we're in just an amazing
1:24:48 time.
1:25:03 Okay, I'm gonna I'm gonna uh I'm gonna
1:25:07 go post this on Twitter because I'm an
1:25:10 idiot.
1:25:17 Oh my god, I love being an idiot.
1:25:22 Why does he always do that? It just
1:25:24 doesn't make any sense.
1:25:27 Oh, I want to get rid of that stupid
1:25:29 [ __ ] thing. Okay. Well, we we we've
1:25:32 got we've got ways to get rid of your
1:25:34 stupid [ __ ] watermark. Yeah. Tools.
1:25:38 Um new.
1:25:42 Um
1:25:48 where am I? Is this it? No.
1:25:52 This it?
1:25:54 No.
1:26:00 Okay, that's it.
1:26:15 Yeah. Let's we'll we'll make this a
1:26:17 little more cinematic.
1:26:24 Beauty.
1:26:28 Good people.
1:26:31 And let's let's go in and let's let's do
1:26:33 some some manual
1:26:36 color correction here, shall we?
1:26:47 [Music]
1:26:53 do
1:27:00 [Music]
1:27:08 d do d do d do d do d do d do d do d do
1:27:08 d do d do
1:27:09 [Music]
1:27:12 Swinging ding ding ding ding
1:27:16 ding ding ding ding ding
1:27:18 ding
1:27:20 [Music]
1:27:22 ding ding ding dang dang ding
1:27:27 ding ding ding. All right. Uh, save.
1:27:33 Now we're going to go to the Twitter.
1:27:36 We're going to put that image there as
1:27:38 if it really happened. Oh, wait. Why did
1:27:40 not not save? Come on. Come on,
1:27:44 computer.
1:27:45 There we go. All right. There's the
1:27:48 image.
1:27:50 Um,
1:27:54 one of the reasons
1:27:57 I was late for late for my AI learning
1:28:04 lab live tonight
1:28:11 was the
1:28:15 Cannonball
1:28:17 fun
1:28:20 gala fundraising
1:28:25 spectacular
1:28:30 that I
1:28:33 founded
1:28:36 and was the
1:28:38 sole
1:28:41 donor for
1:28:47 It was
1:28:49 a
1:28:51 fantastic
1:28:53 event
1:28:55 for the
1:28:57 six
1:28:59 people that showed up.
1:29:05 I invited
1:29:07 [Laughter]
1:29:08 Bert Reynolds,
1:29:12 but he never got back to me.
1:29:22 Um, the the comedy there, if you don't
1:29:24 know, is uh Bert's no longer with us.
1:29:33 All right, there you go. All right,
1:29:36 idiocy. Idiocy.
1:29:39 Idiocy shared.
1:29:46 So, do me a favor. If you would go to
1:29:49 Kyle Shannon onx and uh
1:29:53 big comments or or retweet it or or not.
1:30:02 Oh man alive. All right, people.
1:30:08 I take the train.
1:30:12 Oh,
1:30:14 ran friaking chastic people. All right,
1:30:18 yep.
1:30:20 All right. Um,
1:30:23 okay.
1:30:25 So, the lesson for tonight is like
1:30:33 like I really don't know what I'm doing
1:30:35 yet. I really don't have a sense of
1:30:37 where the boundaries of of this new
1:30:41 image model are.
1:30:43 Um,
1:30:45 and that feels kind of icky. It feels,
1:30:48 you know,
1:30:50 like I think I have it in my head that
1:30:52 like, well, you're on here all the time.
1:30:54 You should know all this [ __ ] You just
1:30:56 can't know all this [ __ ] So, if you're
1:30:59 feeling clueless, welcome to the club.
1:31:03 Um,
1:31:05 but just keep playing, just keep pushing
1:31:07 it. Just keep And I don't give a [ __ ] if
1:31:10 it's this tool or some other tool you're
1:31:12 working on. It could be vibe coding. Um,
1:31:17 just understand that the the more you
1:31:20 prompt, the more that you learn, the the
1:31:22 better you're going to get. And what I
1:31:23 would encourage you to do is share what
1:31:26 you're learning in the AI salon. Share
1:31:28 your images, share your prompts, share
1:31:31 your victories, share your defeats. Um
1:31:34 because we can all help each other
1:31:37 figure out, you know, how to get there
1:31:39 faster. All right.
1:31:43 Beautiful. Fantastic.
1:31:49 Where am I? I don't know. Sleepy Joe,
1:31:52 you're at the AI learning lab and we
1:31:54 were talking about uh learning
1:31:58 N AI and we were uh we just made this
1:32:02 here image
1:32:06 out of
1:32:08 two photographs that were uh part of an
1:32:11 art project I did three years ago. A
1:32:14 photograph I took of Union Station, a
1:32:16 photograph I took of a Colorado sky, and
1:32:18 a Smokeoky and the Bandit Trans Am.
1:32:22 And we just told And we just told
1:32:24 Gemini's new image model, "Put all that
1:32:26 [ __ ] together." And it did.
1:32:28 So, that's what we're doing.
1:32:32 Um, and if you're new here,
1:32:34 I go live five nights a week. So,
1:32:36 tomorrow night is tomorrow's Friday.
1:32:41 Tomorrow night is Friday night, date
1:32:42 night. Uh, you can bring a date. You
1:32:44 don't have to. I can be your date. We'll
1:32:47 just hang out. And then, uh,
1:32:50 >> uh, Kyle. Oh yeah, producer Brandon
1:32:53 >> for Steo tomorrow is Friday, but for the
1:32:56 rest of us here states side it's still
1:32:58 Thursday,
1:32:59 >> right? Because today's Wednesday and I
1:33:01 knew that and that was
1:33:02 >> because you had a podcast with Murphy
1:33:04 today.
1:33:05 >> That was that was really obvious to a
1:33:08 lot of people and one of the things I
1:33:10 like to do on here is see if anyone's
1:33:12 paying attention and apparently at least
1:33:16 at least one person is. Thank you,
1:33:18 producer Brandon. Tomorrow's not Friday.
1:33:21 That's so weird. I thought today was
1:33:23 Thursday for sure, but no, this was the
1:33:25 day I took my kids to South Denver for
1:33:27 treatment. I had the podcast with Dan
1:33:30 Murphy. Yeah. No, today's Wednesday.
1:33:32 Solidly Wednesday, except for Steo who's
1:33:35 in Australia. It's Thursday,
1:33:38 so for him, but it's that's a whole
1:33:40 different thing. It's geography thing.
1:33:44 Anyway,
1:33:50 oh man.
1:33:53 Um,
1:33:55 thanks for reminding me what day it is.
1:33:57 I can't keep up anymore. He's in the
1:33:59 future. He is. Steo's in the future.
1:34:01 Steo prompts from the future. He He's
1:34:04 got He's got 12 hours on us all. Um,
1:34:08 I'm trying to think what else. So, never
1:34:11 mind about that. So, Thursday, tomorrow
1:34:12 night, normal normal night tomorrow
1:34:14 night. So, I'll see you at 8:00. If
1:34:16 you're new here, five nights a week, I
1:34:19 do these things starting at 8:00 p.m.
1:34:21 Mountain time for the most part. Um,
1:34:24 and sometimes it's a bit more
1:34:26 structured. Sometimes, like tonight,
1:34:28 it's like when there's a new tool that's
1:34:30 got some depth to it,
1:34:33 um, I I try to dig into it a little bit
1:34:36 to understand it. This thing's pretty
1:34:38 pretty flipping capable. So,
1:34:41 um, okay. So, a couple of things. I want
1:34:44 you to join the AI salon. So, go to the
1:34:46 salon.ai
1:34:48 and, uh, click on join our community and
1:34:51 do that. But I also want you to check
1:34:52 out, can you pop up are you
1:34:54 readyforai.com? As if, as if he read my
1:34:57 mind. Um,
1:35:00 so we just launched the AI readiness
1:35:04 training program, which if you go to are
1:35:06 you readyforai.com,
1:35:08 you can learn all about it. It's an
1:35:10 amazing program. Um, it's comprehensive.
1:35:14 It covers five different areas. It's
1:35:16 it's not specific to any tool like
1:35:18 midjourney or like if we did training
1:35:22 that was just about midjourney and then
1:35:24 this tool came out, you'd be like, wait,
1:35:26 why am I learning midjourney? Shouldn't
1:35:27 I be playing with nano banana?
1:35:30 The answer is we're going to we're going
1:35:32 to play with different tools all the
1:35:33 time. What the AI readiness training
1:35:35 program is about is about shifting your
1:35:37 mindset. Thinking about how you think
1:35:39 creatively, how you solve problems in
1:35:42 business, how you use AI, you know,
1:35:45 ethically and p with privacy and
1:35:47 security in mind. Um, all sorts of
1:35:50 things like that. So, go check that out.
1:35:52 And uh if you're trying to if you're
1:35:54 trying to up your game with AI, it's
1:35:56 it's actually really important you get a
1:35:58 certificate that you can, you know, brag
1:36:00 about on your LinkedIns uh and tell
1:36:02 everyone you got your [ __ ] together with
1:36:04 AI because I think that's going to be
1:36:05 increasingly important. All right. All
1:36:08 right, people. Fantastic. Bob, tell him
1:36:11 what he's won.
1:36:13 He hasn't won anything. Okay. All right.
1:36:17 Um
1:36:25 [Music]
1:36:26 The answer is always it depends.
1:36:35 All right, cool. Groovy, peace out,
1:36:37 everyone. Uh, be nice to each other,
1:36:39 please, in the comments. Thank you. All
1:36:41 right, I'll see you tomorrow.
1:36:44 [Music]