Transcript · August 31, 2025 · 12.5K words
Ep. 49: Midjourney Token Bashing, Nano Banana, and Weavy with Marco Isle
Captions uploaded with the episode. Expect the odd misheard word. Tap any timestamp to hear it on YouTube, or go back to the episode for the summary, receipts and chapters.
0:00:00Cold open + guest intro (Marco Isle)
Play from 0:00:01 What's up everybody? Midjourney Fast Hours podcast, episode 49, Drew Brucker, Rory Flynn, and we've got a special guest today, Marco Isle Legends only, and Marco's one of the OGs, I would say, also been deep into Midjourney, been deep into AI Image Gen for, I don't know, forever, since I've first seen your name pop up on X, and we're excited to have you here today, man, so welcome.
Play from 0:00:28 Why don't you just sort of give a quick introduction of yourself for those that aren't familiar with you. Hey, thanks very much, Drew. Yeah, so like you said, I mean, I've kind of been dabbling with AI stuff, starting with Midjourney, obviously, the OG for about like three, three and a half years. um And yeah, I kind of just, I started just messing about online and just posting really because there's so many things to discover.
0:00:35Marco’s background: design → UX/UI → fashion
Play from 0:01:01 was kind of just posting like an online journal for myself and then people were kind of digging the stuff and it kind of just snowball from there. You know, like I always try and stay true to what I like it rather than like going with, you know, just noise. Like, so if I want to do like fashion stuff or whatever it might be, I kind of just stick to that, you know. And then I just post and share findings because obviously there's a million things that are possible and it's just nice to kind of put it out there.
Play from 0:01:30 If it can help other people, great, you know, and that's, that's kind of it really. What's your origin story too with the interest in AI image gen? What's your background like that maybe connects some of those dots? Or was it just totally different and you found this and you're just like, this is a brand new thing that's something I didn't know I cared about? I think so. So I've got a background in like um graphic design, user experience and interface design and fashion as well.
Play from 0:01:59 Like I basically designed baseball caps and t-shirts and used to like have several websites like selling third party um brands, you know, online. And I actually got into like the digital space just by setting up a shop on eBay back in the day. And I was annoyed at how I was annoyed at how shit the listings looked when you, when you added your products. I was like, okay, if we make the listings look nice, can I charge more for said products?
Play from 0:02:33 Or can I get more? I'm it's like going into a store that's like peeling walls and all the rest of it, unless that's the vibe, but like, you know, you're, you're, you're a bit sketchy and you're a bit unsure about. Part of them, we do money, whereas if it's like really nice and wrapped, it's marketing, you know, it's one-on-one. You can have, you have a really shit product and market it amazingly and do really well. You can have an amazing product and market it poorly and it doesn't sell, you know? So it was kind of, I always love, which is why I kind of got into UI and UX.
Play from 0:03:01 I love aesthetics and mechanics, how it looks and how it works. And I think they go hand in hand. Oh, love that. We talk a lot about a mechanics on here with with the AI image gen, right with the prompting Sort of like mindset you have to have especially early on with the models, right? It was almost like it was hard to get things without describing them mechanically So I think that's an interesting. Yeah word that you use there. Go ahead Roy Now I saying, Mark, I've got a question for you.
Play from 0:03:28 On a scale of like 1 to 100, what's your creative ADHD levels? I'm uh pretty high up, pretty high up dude. I think we're all sort of in that. I think that's why we like find each other, because we just know that like is or anyone that's basically in the space, right? Like if you don't have some level of creative ADHD or like curiosity, I don't think you know, for people who don't know Marco or haven't been following him, he does great stuff on AI.
Play from 0:03:59 But some of the other stuff that you've done is super cool, too. Like I remember you putting together some magazines and stuff like that. you know, product design, your design and t shirts, hats, things like that. Also, we just talked about this before we got on. Marco likes to cook a lot. So you know, some of the things that he's dreaming up in the kitchen, that's just that's not because we have AI tools. That's because we have creative, you know, we have a creative on here and a creative mindset.
Play from 0:04:24 So dude, I just like I said, I really like how you sort of like bridge all of that energy. out there creatively and put it into the AI work too, because you can see it a lot of times. I remember some of your stuff early on that I saw first was like the, it was like fashion and like the Greek Acropolis with like the, you know, the graffiti and it's like, there's more layers than just like, me a woman in like a fashion pose. There's like a lot of intention behind it.
0:04:48Aesthetics vs. mechanics; the cooking/technique mindset
Play from 0:04:51 Yeah, well said man. Yeah, I think there are there's so many depths to what we do and any kind of like everyone. Everyone has really interesting backgrounds that they draw upon to like bring into the mix when they're kind of like exploring. And I think, you know, we all know that Midjourney and other image gen tools, but mainly Midjourney, because obviously it's the OG are all visual things, right?
Play from 0:05:17 So you I mean, when I'm when I'm kind of doing stuff. I'm not just typing stuff in and then iterating that looking at what I'm seeing like a season. I like all that. That's cool. I didn't mean to do that, but actually that's quite cool. I'll start like going off on that direction. It's just it's a visual thing that you can kind of just navigate and it's really nice. You know, it's like it's very similar to cooking. You know, you have a recipe like that you follow but it's more about technique because if you know how to make a cake.
Play from 0:05:49 Right. You can change the flavors. It's literally that like once you know the technique and the process, the structured way of organizing prompts, can then swap out the angles or the parameters or, know, the lighting or whatever it might be. Hey fellas, really quick, Rory, let me know what you want to do here. I've got a bunch of tabs pulled up that I wanted to share his stuff on, but my Arc browser crashed this morning, so now I'm using Google Chrome.
Play from 0:06:20 And of course, I've got to close it to be able to share. So I'm thinking I'm going to jump really quick and jump back in here so I'm able to share, because I'd love to use the visual to speak to while he's talking through some of this stuff. So short story, just hang tight. Give me one second. short story. Short story. So the Nano Banana stuff, I saw you post something recently, your flow that you were showing me.
Play from 0:06:49 So cool. We're definitely going to talk about that. We got to get into it because it's too powerful. Putting it together in the keyframes showed me how good it actually was. There's not even morphing on a lot of the details of the car. So it's almost one to one, which is insane. car or was that like I just prompted that car? That was a Midjourney image, and then just took the angles and then was like, all right, let's keep rolling with it.
Play from 0:07:17 video showing the roll around is like so crisp. It's insane. It's, you know, so where are we starting back from? Back. going to I'm going to cut it from that answer. What were we remind me what we were just finishing up with there? Talking about Mark said something about cooking and the ingredients and like, say if you know how to bake a cake, it's the same thing. So.
Play from 0:07:42 going to go ahead and pull this up and then we can just jump into something here. That's right, let's roll the tape. There we go. All right. So we're just taking a look at Marco stuff here. Roy, you were talking about some of the stuff that he's put together. So he's got some of stuff here, right? Marco, you've got sort of like, I guess, some prompt collections, but you've also got the prompt magazine stuff that you were doing too. Is that different from this?
Play from 0:08:08 so the prompt magazine was a collaboration I did with Free Pick um for the first upscale conference that they did in Malaga. So we just did a magazine as like a little goody bag thing to go with the show, which was a nice thing, you know, because like in this digital age, it's so nice to actually be able to have something tactile to flick through.
Play from 0:08:34 It's great, man. That was that was really cool. I remember you showing me sort of like the layout and stuff. And I was like, this is like this is where it looks cool and hard copy. It looks it's not you know, it's like getting Midjourney magazine or or the you know, like the special edition book they put out but version. Yeah, that's great. Yeah. Summit day. an extended version via PDF, which I can send to you.
Play from 0:09:01 Yeah, and if there's a link to that, let us know. We can include that in the show notes too for folks. Yeah, I was looking at this. You've got this. And then obviously, good follow on X. That's your primary channel, correct? Yeah, like to be honest, like when I saw next, wasn't, I wasn't really posting on LinkedIn until like probably earlier this year. Um, I've just been kind of doing stuff on X. It's so funny, like looking at my feed in white, cause I've, have my, I've got a dark mode on my screen.
Play from 0:09:30 Yeah. It pops so much more on the white. m that's what I'm like switching browsers all the time and it always defaults to white. That shot there, Drew, I started doing a whole lot of thing around mythology and that mythology meets a metropolis, like really cool, like kind of like all these characters thrown into like New York or, know, just quite cool concept.
Play from 0:09:57 like the dark gritty vibe too. Yeah. What I like about both of you is I can always look at the images and know exactly who it's from, which is like after, you know, there's like a, there's a lot of exploration and stirring, like, know, straying from the mean, but like also at the same time, like I know a Marco image. I know a drew image when I see it. Like if I didn't see your handle, I'd know it's you. Cause there's like a, there's consistent themes just like sort of baked throughout a lot of the content.
Play from 0:10:25 And I know it's like, you know, you a lot of the stuff that you do and the angles and sort of like the exploration. And then Drew is like, until right away, like I know Drew, I know a Drew image out of nowhere. Like you could, you could put that in a lineup and I'd be like, that's true. 100%. But it's cool. Like the metropolis, like that, that mythology metropolis, seems like a lot of that is sort of baked into what you're doing overall. Is that, is that something like what, why that? Well, so going back to what you were talking about previously, like the graffiti kind of like Roman columns and stuff.
0:10:50“Rome Reimagined” + early Frames testing
Play from 0:10:59 I basically, when frames first came out, I run my frames. They asked me to do some like testing prior to it launching. And I really liked the idea of like, I called it Rome reimagined. So it was like taking all of the elements like I can think of from ancient Rome. obviously the architecture, the clothing, the environment, you know, all that kind of stuff.
Play from 0:11:25 And actually then bringing it into like 21st, yeah, 21st century. So, and I did like MMXXV, it was the thing. So was like 2025. And, you know, like looking at, I did a, I did a whole bunch of like plinths with marble sneakers and stuff on them instead of like Roman busts. and like graffiti columns and then looking at like having the togas but having it like um camouflage for example instead of just normal silk just really experimenting with that and then yeah like i love that that was another one actually the the girl with the horns you uh if you go that's part of it yeah if you click into that thread um so she's got like golden horns basically yeah that one there um
Play from 0:12:14 Yeah, I kind of just started doing these. I've got a really cool prompt that I was using. And yeah, man, I made a few mood boards and then just like this prompt seemed to just open up all these really cool things. Yeah. m we start to petition Midjourney to get multi-prompting back? I know you like that. I don't understand. know they changed their architecture right.
Play from 0:12:40 For when they did a V7 onwards, but I actually did my last PDF series on a gum road is actually on prompt stacking or multi prompting. And you can really get some interesting stuff because you're, you're kind of fusing multiple ideas and directions and then waiting them in different ways. like. You know, obviously you guys all use seeds and you fix the seeds to test things, but like just fixing the seed and then like waiting things one way or another when you when you have two or three stacks.
Play from 0:13:14 So interesting. I really like it. You can get some amazing stuff. I don't yeah, it's interesting to like to hear and wonder what other people do inside of Midjourney with with waiting in general because I Have always found that fascinating You know with with image and style rough to you know, just like how you can dial those in uh You know and just get you know completely different results or even just blending right in waiting and you know It's just there's so many different variations of a single uh Idea or or sort of like workflow within Midjourney to get unique results just by dialing in the weights.
Play from 0:13:54 I mean Yeah, it's it's it's really pretty wild What's your, what's your like, here's a Midjourney question. If you're going in there, what's the, what's the feature that you're using the most? you using mood boards, SRF, multi-prompt, waiting? I actually, again, I did a, did another PDF on like how to recreate S S without using an S ref just using natural language basically.
Play from 0:14:22 So, I mean, I would, I don't use that much. Like obviously they're very powerful. Like when you use an image as a reference or whatever, but yeah, I kind of just like seeing what I have from my, from my prompts and I actually I built a tool, I think I've, I've run some stuff that you sent me Rory through to like extract the tokens.
Play from 0:14:47 So yeah, basically I was, I was, it's like the fifth version of this tool I've been like playing with. It's like a custom GPT basically drew. And, um, yeah, so like, obviously it does like, you can, you can upload an image and it does like a, it gives you a decent prompt and then like a structured breakdown as well. And then I was playing with like this experimental thing where you like fusing some of the words together because obviously an LLM doesn't look at text how humans look at it.
Play from 0:15:15 So I got some really, really interesting results by doing that. And it kind of gives me five, like sent like very short sentences fused together as one prompt. So I can run that as a whole thing. It's almost like prompt stacking without prompt stacking, like without waiting. Um, and then you can like take out each individual statement and then run those as well. you get. other stuff just endless endless mate though.
Play from 0:15:41 it's, really like, that was the first thing that I noticed with it when I was, when you were sending me the prompts back and it was like, you know, I was getting these, these crazy like term, it's like term bashing, right? It's like almost like blending two images, but with like blending two tokens. remember it being like getting something that was like a, like a Jaguar or something I sent you. And it was like, like Apex integration or something is like one word. And I'm like, I don't know what that means, but it's cool. And it worked and it gave me some really crazy stuff.
0:15:59Token bashing & Marco’s custom GPT flow
Play from 0:16:08 So Yeah, we we've talked about a bunch of prompting techniques on here, Drew. I don't know if we've ever mentioned sort of token bashing before. No. and I was that's kind what I was looking for because I know you posted something on that Probably was a while back um So maybe we can pull it out or not Do you is it is the GPT something that like is far enough along that you want to show here?
Play from 0:16:36 Or is it still work in progress for you? mean it works fine but like it's just not public but I can show you if you want like how it works. Yeah, what's... I love these too, these are great. That's actually with rev. Even though I wanted to say Reeve. That's how I always say it. No, I always say Reeve because of the E, right? But it's actually rev apparently.
Play from 0:17:02 Exactly. right? Like these sort of words that just are two words that form one word, right? It's this net new word, right? Yeah, it's kind of dope. I've never really played with that concept before, but I think that's a really cool thing. ah so, know, like, Rory, we've talked about a lot of things on this podcast to the point where it feels like we're always like, we've talked about this before, but this is, uh I think this is sort of like one of those novel things I haven't seen before, so.
Play from 0:17:28 of the one of the other things as well, this is a while back but I was I'll if I can find it but it's because it was quite a while I think it was last year, but Literally some of the words that we I was just trying to work out like illumination and enlightenment or whatever like like putting them together so like Illuminatement or whatever all these kind of different things and some of the results man were like insane like I did a like a provocative Forest of thing so it was like I can't remember but was like provocation but the end bit instead of occasion was like uh illumination as well like there's like I know it's just light and eroticism and all sorts of cool stuff uh it's really interesting I find that kind of stuff it's just it's untapped I think David was saying from Midjourney he was basically saying
Play from 0:18:24 that not a lot of people use the advanced stuff like all the weighting which is why they kind of weren't too bothered about it but Yeah. But it's fun because that's the creative piece of it. That's not the control. That's the like, let Midjourney be Midjourney, do its own like wacky stuff and you end up with really cool things. Cause what like the creative part on both sides, right? Because it's like, that's not really something that Midjourney would have thought to do or promote.
Play from 0:18:50 It's something that you thought about that's this sort of like new idea, but then Midjourney can execute on it creatively as well. So it's kind of like this combination, which is pretty cool. Because it still is the only tool that, to me, that allows you to be that creative and not have to force control. You know what I mean? Like, it's not about force control. It is about just like, let me smash two random words together, and maybe this will make something cool.
Play from 0:19:16 I do love, love, dude, that was like one of the, when you showed me that, that was the first time I had seen anyone sort of thinking that way. And even for like, instead of it being, like even in the prompt structuring thing, so like, again, for anyone who doesn't know what we're talking about, like, the GPT that Marco built, like when it spits out the prompt, like, you know, it's in a structured fashion, but instead of it saying like subject and then like colon and like this, like I saw it up in one of the things before it said like Metro Muse, like that was like the name of the subject, which like instead of naming it as subject or environment or whatever, it was like utilizing like a token as like a holding category, which I was like, all right, cool.
Play from 0:19:54 That's another little like micro tip. Yeah. Like is, the thing is with that tool, like, so you can, you can basically put in like a Pinterest image or whatever it is, like a reference image you like. But then I was kind of like putting my own images in. like kind of just continually evolving and just going down a complete rabbit hole with like seeing what would, seeing what uh prompts and tokens it would give you from reinserting images created from the prompts and tokens it given you previously.
Play from 0:20:30 It's like insane. Okay, so I know we got a lot to talk about today, but I am curious, you know, like you've, to me, you're always like talking about tokens. And I think like this, you know, token bashing thing adds another level to it. Do you have any general favorites that you like at the moment or that you're experimenting with or that you just always go back to time and time again?
Play from 0:20:59 I mean, the one you actually did a post the other day, which I replied to about shadow play. mean, that's one of my favorite things. Rory loves a bit of shadow as well. And it's just awesome. Just that really like high contrast, you know? I actually did a whole PDF on shadow play as well. And just playing with the different types of shadow, you know, there's like, I love dappled light.
0:20:59Favorite tokens: shadowplay & moonlight
Play from 0:21:25 You know what gives amazing lighting? Moonlight if you put in moonlight as a token it's insane. It's really enough Have you tried Gobo Shadows? Gobo Shadows is a good one. Bo Shed. Open up Midjourney, run Gobo Shadows. I need to know what that is. I have no idea what that is. shadows would be like something like this.
0:21:32New token: gobo shadows (why it works)
Play from 0:21:51 Okay. So it's like it's an object in front that's got the shadow. Right. So uh let me see if there's like maybe. Yeah. Like go both shadows. You know. Simple prompt. was I was actually like before this I was like what are you know what are some like tokens that I haven't talked about on here before. That was one that I that I like.
Play from 0:22:17 uh mean, you can use that for, obviously not for portraits, but I'm always thinking in a large way just with like portraits. um think the other ones that I always, there are a couple like really cool ones recently that I found like pixel smear and data mosh effect. So I think this one, I think this is a data mosh effect.
0:22:19Glitch looks: pixel smear & datamosh
Play from 0:22:45 This is. You know, and this is not a complicated prompt, right? mean, this is like super simple stuff. But I really like the effect of stuff like that. um We've doing quite a lot of glitch stuff as well, with, like, but like you try fragmented bodies. That's pretty cool. Okay, yeah, let's run a few like random things. I always like to like you reminded me of one the other day, which was uh chromatic aberration.
Play from 0:23:15 I just think that's always a good one. ah We could do, let's do some other ones. do, oh, another one I run a lot is like just weird core. Like when I'm doing like the style reference, the style code exploration, weird core is kind of a cool one. ah try prismatic rainbow light streaks. Okay, I like where this is Prismatic rainbow, alright.
0:23:34Prismatic rainbow light streaks (stacking ideas)
Play from 0:23:47 asthmatic rainbow. Yes, this is what we're talking about. This is the sauce. Yeah, mean, like, there's, I mean, there's just so many good, there's just so many cool things like this where it's like, if you throw that in there, changes the entire image, you know? So. do like... What was the other one I had? mean, like colors and stuff you could do em pale monochrome palette as well is quite cool.
0:23:54Androgynous beauty as a control aesthetic
Play from 0:24:19 I really like, m I use androgyny a lot as well. I quite like how that is like neutral androgynous beauty. See, there we go. That's another good one. course, I never know how to spell it, so I always have to do that. Yeah, dude, there's a lot of good ones.
Play from 0:24:47 I uh just found this prompt from Marco that we used a while back and there's so many good ones in here. Just even the token bashing and the random ones, I would have never come up with this. What was that other one we just saw that was on your feed too? was something, Metro Muse. I mean, and look, like this, think we're just gonna run this token, right? So this is not in combination with anything.
Play from 0:25:13 This is, guess, more or less gonna just show us kind of like the results of what we're talking about here. ah Using EXP at all, Marco? Yeah, I do. But I find lower ones give better results. I find the higher you go, kind of just like makes it very almost like stipple glazed, very dotty. I don't know if you've seen that or not on your experiments, but I tend to do like three, seven, 13 and then like 20, 27.
0:25:22exp sweet-spots (3, 7, 10, 13)
Play from 0:25:46 Welcome to the podcast. We're in it, baby. This is it. This is the stuff. Fragmented bodies and of course like so this is different too because it's got my personalization So this is all this stuff is gonna differ slightly for everybody, right? uh But I think this will be kind of cool just like look at the like this is like crazy Yeah, I'm really finding great.
Play from 0:26:12 Yeah, like between, I would say zero and 10. Like just breaking out of this box of no exp, I typically almost always like the four grid better with some exp sprinkled in. Look at the that chromatic aberration. Always works. it? Wow. that's a, there's some, you know, again, there's some goat horns.
Play from 0:26:40 Prismatic. All right, so now we know we can, we can add here. that's imagine that stacked so that prismatic rainbow lightstripe stacked with like cinematic fashion or you know Go-go shadows with a prismatic rainbow light streak. Yeah, there it is. Like that. Pixel smear, pale monochrome, palette.
Play from 0:27:06 That's dope. at the texture on that flower. That's, yeah, like, hey, you know, this is what we do here. That's right. I did a whole load of uh religious kind of stuff, but like cool religious, not like, you know, in your face. It was, I've never heard this term before, liturgical drift.
0:27:17“Liturgical drift parade” prompt set
Play from 0:27:32 Whoa, okay. Let's. And it was like, was like priest and I did it with like motion blurs. It was like priests like floating with motion trails and stuff. It was pretty cool. Let me see if I can find that. Look at this. Then Metro Muse. Yeah. Yeah, it's cool. there's some, there's some fun ones, even just, uh, I was looking in here. There's so much I have.
Play from 0:28:00 you so much random tokens from this prompt. Should I share it? Just to show the actual prompt itself, just to see how ridiculous it is. Here, this one. Because I was digging this one when we were using it. It was just like... And it was a, let's see, layered composition, text box, levels of the shadows.
Play from 0:28:30 gaze. Right? Shadowhunter, um Apex Glare, that was a good one. Ritual Fang, like their editorial beast. Primal, you know, we have the Ritual Pro 5 Hasway medium format. Unparalleled Precision, I sort of like that um combination. I use High Precision a lot, but I like Unparalleled Precision too.
Play from 0:28:57 But there's so many, like Tribal Iconography, Obsidian Etched Fur, like Etched Fur is a good one too, because a lot of times that's just like, or Velvet Depths, Feral Whisper. Like these are, I love that kind of stuff, because I'll play with that all day. Obsidian Sculpt, you know, like these are, it's fun. It's fun to play with this. Mid-Journey Coherence is special.
Play from 0:29:24 ah They have a little bit of that magic just sprinkled in there. Where if your model's too coherent, just kind of, it's not gonna do anything. Yeah, yeah, it's pretty insane. But I'll stop sharing this one now. That was just like, I remember that prompt specifically for having sent me on a rabbit hole with those words and then just ripping. It was like the multiverse in there.
Play from 0:29:49 It was like, it's taking one word and going this way and this way and this way and this way. Yeah, it's really fun to play with on the token side. I still think that's the biggest exploration. There's so much to be unearthed. Yeah, like, you know. sent in the chat that liturgical drift parade, the prompt and also the link to some of the images I created with it. It's like literally one of my favorite image set that I've done. It's so nice.
Play from 0:30:15 I'm gutted at any goal. I don't see how it would. No views. Maybe I did it too late or something. What aspect ratio you think I should put this in? 4 5? I did 16.9 actually, but try it. four five. That's another thing actually talking about aspect ratio, literally just adjusting, you're doing permutations with aspects like the same prompt just changes so much depending on the aspect ratio you use.
0:30:24Why Aspect Ratio is everything
Play from 0:30:51 It's probably one of the most important parameters, aspect ratios of the whole thing. 100%. Yeah, there's definitely uh biases baked in based off of that and the training day that's in it, 100%. Yeah. And obviously like if you're doing like a cinematic shot, should be probably more landscape and versus like a close up should be, you know, four or five or whatever.
Play from 0:31:17 yeah, so here's your results here through the, yeah, sick, yeah, vibey. So nice. Now Drew, if you had a gun there, now we got a whole crossover between the two. None with a gun, baby. You want to know which one I've been liking a lot aspect ratio wise? 1-2. So I know a lot of people don't generate in 1-2, probably, but it's been really good for like cropped vertical composition, like really solid.
Play from 0:31:51 asymmetrical, unbalanced, like, no symmetry at all. It's... I love it. as well at one point. Yeah, it's just taking the wider angles and just making them... It's like 916, but 916 sometimes is already almost too structured based on probably the data that's input from 916. So it's like, let's make it thinner and let's see what it comes up with. It's great for closeups. It looks awesome.
Play from 0:32:19 So I ran this with some mood boards on. It's cool. Yeah. Yeah, it's interesting. is, this is sick. one there top left as you're on your screen now it's really cool. Yeah. Yeah, these all have a story inside of them.
Play from 0:32:46 It's very cool, man. love I love motion. I really do. That's probably one of my favorite things. Motion, shadow play. And then just like I love the juxtaposition of things. like the whole Rome reimagined. So you have like this ancient Rome. In fact, one of the prompt stacks I did was like I took um some of this kind of motion, know, em light, light streaks and all that kind of stuff. And I did a Roman soldier with light streaks and then like PVC avant-garde editorial fashion as different stacks and I waited them differently and like you get like PVC busts of like Roman soldiers and then like cool stuff man.
Play from 0:33:27 Really cool. Hey, tell me this. I think probably people are curious and I'm curious too. We all probably have different ways we go about it, whether it's intentional or not. when you're, say that idea specifically, right? What's the trigger for that idea? Are you um finding something with, is it as simple as you finding something that's Roman and then all of sudden you've, why not pull this into an urban environment and then it starts to snowball?
Play from 0:33:57 Do you see, uh a particular token and something that you've created before or inspiration from a past image and you're like, what if I use this in a different type of environment? Do you use LLMs to brainstorm? Like what is your creative building process look like for idea generation? So many layers to it. So when I was younger, I used to literally just sit and mess about with Photoshop and layers. I used to make my own little look books with fashion staff and just messing about with all the patterns and that kind of stuff.
0:34:12Idea sources: mood boards + Cosmos.so
Play from 0:34:30 So always be really interested in colors and that kind of vibe. And yeah, we all love looking at visually pleasing images. And one thing, mean, obviously a lot of people use Pinterest, but there's one that I go on. It's called Cosmos. don't know if you've heard of it. Cosmos.so. Really cool. It's basically like Pinterest for designers.
Play from 0:34:57 Rory, you'll absolutely love it, mate. It's so good. It's insane. got a good it's a really good curated mood board type feel to So like I've got a lot of influences from the stuff I was doing previously, like with the fashion stuff. I've always loved camo and I, and I always like graffiti and streetwear and that kind of stuff. But I think, yeah, like I think when you see, when you see a particular image, kind of like, sometimes it gives you an idea or like sparks some kind of inspiration and like even my own images as well.
Play from 0:35:33 Like I'll kind of iterate on them and say, that'd be those two things together. We actually pretty cool to put them together. And then, and then I had this whole, yeah, that's actually really nice as well. had, had this whole thing of like trying to mine the perfect prompts basically, which is why I started messing about with like custom GPTs and just documenting stuff, you know, and trying to see like, how can I improve my process?
Play from 0:36:00 Like obviously. We all can prompt and we all have ideas of doing things and you can tweak the parameters and stuff, but how can we elevate that by diving into not only the work that we create, but then looking at influences and then putting those into the GPT, extracting tokens and then messing with them. It's infinite, right? It's an iterative process that will probably never end.
Play from 0:36:28 just an ongoing creative journey, which is... is awesome and like it's like fashion in itself and general style like everything recycles you know like stuff that was trendy in the 80s now becomes trendy again in 2025 and you know people pull like animal prints into fashion or whatever it might be you know so but I very much like I probably approach in education the same way I approach food I work with my eyes you know You kind of see what's cool and then you kind of go with that.
Play from 0:37:04 the ingredients on the table and then just be like, what can I do with this? Right? Like that. feels like a reps thing too, just like with most things. It's like if you're doing this, um the more you do it, the more ideas get unlocked, right? Because you're kind of breaking outside of the box over time. You're finding more more ways to utilize this or do that. And it just opens more doors and you find different pathways to go down. Versus if it's more of an infrequent thing or you don't do it enough, those things just aren't gonna click.
Play from 0:37:35 Exactly. There's some. met that was really cool. They there with the NFL guy that you just went past. Yeah, that's Are you running a lot of this stuff uh on Midjourney video too, like when you do it, or is that just like uh an afterthought? I have done some stuff. To be honest, I've run some stuff. I forgot I run as videos. And then I went back. was like, shit, that's actually like insane. um You get going in that prompting hole.
Play from 0:38:05 And then it's just before you know, you have like, I'm just generated like 1500 images and then like bang, like I did like a hundred videos in between, but I forgot about them because I did, you know, 700 images after I ran it. Definitely, definitely done that. Yeah. some more tokens. m Motion smear, that's a cool one. You know, just cool stuff, guys. I mean, I think people love tokens too because I think, again, like you can utilize one token and that could generate and spark an entire flow or new idea that maybe somebody jumps into.
Play from 0:38:38 So I think that's really cool. Haven't seen... Yeah, yeah, haven't seen that. NFL me do it a our RGB spectrum trails nice. I like this Some cool stuff in there. while we're, mean, look, there's so many image models, video models now. mean, updates change, they're coming every week.
0:38:54Nano Banana enters the chat (workflow + why it’s different)
Play from 0:39:05 Maybe the prioritization of these things change. What are the tools that you're finding yourself using the most right now on image and video? Maybe like. Is it again, of course, like maybe it's just use case based, right? What is the outcome of what you're trying to do? Is it for play? Is it for more of a client? You know, like maybe that alters it, but like, what are you, what are you sort of like falling back on at the moment? I think, like you said, it's dependent on what you do, right?
Play from 0:39:32 Like, I think my default is just like Midjourney for exploration because nothing even comes close to it. it. Never will. It's just game over. But like, so the stuff that I did, uh like that Metropolis mythology stuff, the original images were created in Midjourney, but then I pulled that into NanoBanana, right? Which is this whole new workflow. So that's actually something I really love doing.
Play from 0:39:59 em And I did something else that I showed Rory, which I haven't shared yet, but it's basically called where petals fall. it's like, the idea is like, it's this volcanic kind of wasteland, really black with like black soil and like a few spindly like yellow wildflowers. And instead of raining normal water, it kind of rains petals. And then this woman emerges from the soil, this black soil, and she's wearing like a yellow, bright yellow.
Play from 0:40:26 dress that's kind of in the same vibe as the petals and the flowers. And then she has this kind of like romantic exploration of the landscape and she kind of navigates her way up to this like stone tower and she's kind of like shots of her hand on the stones and withholding the bouquet of flowers. And then she kind of like climbs up to the top and holds them up to the stormy sky and kind of lightning strikes and then the petals fall and it just kind of goes black.
Play from 0:40:52 So that's my vibe and I... I did some shots on Midjourney and then I just created literally like about a hundred like insanely dope shots in Nano Banana. And like I was doing them in LM arena, which was painstaking, but because it was so, because it was so good. Oh man. Although I think did get a good run. feel like the more times you selected Nano Banana in a session, the more it was shown bizarrely, which was pretty cool.
Play from 0:41:17 But yeah, I only, I only stuck at it because it was so good. The results I was like, I don't care. I'll wait. But then obviously when it got dropped into FreePick and all these other tools, I literally did about 100 shots in like 10 minutes. I was like, shit. Yes. feel like this is the perfect segue then, because we knew we were going to talk about this. We knew we were going to get into Nano Banana. This feels like a good time to do that, right?
Play from 0:41:43 uh I know we've all three been playing with it. And some of the most recent stuff on your X-Profile showcasing that too, right? So uh let's jump into this, right? ah Now that it's not just accessible through LM Arena and we can access it directly. I mean, How is this changing the game for you? Maybe from a workflow perspective and maybe also just from like a what's possible perspective.
Play from 0:42:15 I mean, to be honest, again, like I shared, don't know if you saw it or not, but I did like a cinematic reenactment for the book of Jonah with PJ about a month or two ago. And we were doing that from like January till April, May. It was like 42 scenes, 15 shots in each scene. It was like, I probably made like four, 5,000 images minimum on, I actually use Google Imaging for that.
0:42:15Case study: Book of Jonah (42 scenes, ~5k images)
Play from 0:42:44 because it had pretty good consistency but even that like we had to impaint like so many shots and so that workflow like literally if we had Nano Banana would have been cut by 10, 15, 20. I could probably do that in a day mate I reckon. want that hour of your, those hours of your life back instantly. But then on the flip side of that, know, like I love attention to detail. I, was, was kind of annoying me that it wasn't perfect.
0:43:03Video models: Veo-3 notes + start/end frame
Play from 0:43:12 So I was trying to like, you know, get, get it good. And, it's just crazy. Like in that period of time, how quickly the tools changed, like from what we had available, like Veo three wasn't even released, man, like text to video. We actually did our whole prologue for that uh video for that film with Veo three text to video. But yeah, it's just. It moves so fast. think you obviously have like a core tool set that you go to, but it would be silly to just sit on your laurels because I mean, like Rory's been doing with Weavy, like some of the stuff he's been doing, automating processes, because I think that's where it's going now.
0:43:33Weavy & automation; setting guardrails
Play from 0:43:50 Now that pretty much everything's possible, especially with Nano Banana, it's not really, oh can we make this? Yeah, you can. You can make whatever you want, but what do you want to make? And it comes down to more like people's style. and their taste, think now more than anything else. Obviously, stories really important throughout always has been even not even AI stuff just in general. em But yeah, I think automation and all the tools we have available, you kind of support for choice, aren't we?
Play from 0:44:20 yeah, you almost have too many options, right? So many. it's, you know, I think that's where that's definitely where I find myself. I put this in my last like stage presentation. I'm like, I showed like a little diagram of where each thing could go and what you could change. And it was like, you know, this image can go here and then it become another image or it could become a video or it could become a 3d asset or it could become a piece of code or it could become text. Then from there, you can change this. And this is the I'm like, that's like the infinite multiverse.
Play from 0:44:48 And that's where if you're don't have some like guard rails or some sort of like limiting factor, you're just gonna you're gonna end up not getting anything done because you can go. Yeah. like, shit, what did we just spend the last hour or two doing, right? We need some guardrails. I've talked to companies about this too. I'm like, to people that are implementing these workflows and these pipelines into their, into their agency, into their brand, into their business. I'm like, you know who you also have to train?
Play from 0:45:14 Not the creatives, the people at the top who are making decisions. Cause if they just keep saying, well, what about if we put it here? What about if we change the color of the dress? What about if we do this? And it's just like, they also need to be like refocused. Like no million, because you can make a million edits doesn't mean you should make a million edits, right? Yeah. because that's the bane of all existence anyone here listening from an agency, you know, point of view. If you know what I'm talking about, like, what about if it was a little bit brighter outside?
Play from 0:45:43 What about if the guy was wearing earrings instead of a necklace? What about if you know, it's like, too many because it's all possible now. So we have to like, or else nothing's gonna get done. Like, I'm sure it's crazy, though. I'm sure, Marco, you felt a lot of that with the with the Jonah project. uh Like some of the shots that I made, I'm genuinely like really happy with. Like they look so cool.
Play from 0:46:09 There's a shot where basically, I need to see if I can find the actual thing. em But basically there's a shot where the whales kind of like spat Jonah out back into the ocean and the fishermen have noticed it and they kind of like go into the water and kind of like cradle him. And there's a shot of. two fishermen holding him and he's like limping their arms. It's just such a nice shot. Like honestly, when I made it, I was like, wow, that's so cool.
Play from 0:46:34 So like for context, Drew, we had like a proper script writer and a trained GPT on so that we had the script. had a custom prompt that I built. We had about 10 different characters plus all of their outfits. We had about five or six locations from like the King's bedchamber to like the seaside, the port, um the ship deck, the ship like quarters underneath.
Play from 0:47:07 That's incredible stuff. looks, I know this looks bad now. But that's the crazy thing with longer production cycles now, like for these longer pieces that people are doing. It's like, if you spend a month and a half on a project, like by the time you're done, you're like, I want to go back and redo the whole thing because it's so much easier now. Right? Right. Yeah. at the same time, you know, that's where it's like, man, when you talk to other people in the space, you feel a little bit less bad about it because everybody knows that.
Play from 0:47:42 But like, you know, when you try to describe to somebody that's not, that's where it kind of hurts you to say, my God, like this tech is old now, you know, but this still looks really good. Yeah. you the post with the shot I was talking about. There's four shots that I really like in that. This is, this was, when did you guys release this? This was July. That was, but you guys were working on that since when?
Play from 0:48:10 That was a couple months. Yeah. Yeah. I know that. right one, the bottom right one. So sick, love it. So for anyone who I know is not listening to this podcast, who might be listening for the first time and think that AI has a magic button, there you go. Six months. That's a typical production schedule. or five thousand images to start. That's just your portion of it too, right? I mean, that's like a lot.
Play from 0:48:36 That's the same thing as reshoots. It's same thing as... It's all the same. It's just a different medium that it's capturing the stuff, right? Yeah, I found it. mean, we actually built out with Imogen, like all the characters, like literally prompting, I prompted like dry linen, wet linen, all these different things, you know, just to like get the different vibes for the outfits and costumes, different lighting.
Play from 0:49:06 are great stills. I love this eye. That's great. yeah. Basically the king is having a premonition and that's like the whale's eye. um sick. It's awesome. The biblical content is like so prime for it. It's like so perfect for AI, you know, like it's because it's it's also like how you want to interpret it. Like everyone's got their own interpretation of the stories anyway. So it's like just visualizing it in a way that looks cool and might be different.
Play from 0:49:34 Like you guys did a good job with that. the funny thing is PJ did this, like, I don't know if you saw it was like mocking all of the biblical stories. did like a something that went crazy viral. It was like Jesus in the tomb waiting to get brought back to life from God and stuff. I know if you saw it, it's really funny. It's so funny. knack at that. He's gonna he's gonna knack at that stuff. But it's crazy man.
0:50:00NanoBanana pros/cons; real-world limitations
Play from 0:50:00 this is, mean, so let's circle back to Nano Banana specifically, right? ah Maybe some of the folks listening to this have already played with it, maybe some haven't. What are the things, like let's go to the things that are really changing the game with this. Why is this such a big deal, right? ah And then let's talk about sort of like maybe even some of the limitations that we're currently seeing. I think that might be a good place to go.
Play from 0:50:28 Totally. There's so much. I'm jumping in on this one because I've been trying to play around with context, Flux context, and runway references, and ChatGPT, and they all have their own limitations. Flux, to me, is flat. I don't like Flux because it looks flat. There's no depth. It's really hard to bring that or take that saturation down and take that AI look away from it.
Play from 0:50:55 To me, Flux has always been like, utilitarian, but not great. Runway, a little bit better, but it also sort of like has its limitations in terms of what it can do. um And then there's, know, chat GBT, you can't do 16 by nine, you can't do, you know, nine by 16, you have to do whatever the aspect ratio is, they're all three, two images. you, yeah. And it's, and you know, it can be really good, but it can also just be really bad and flat again.
Play from 0:51:24 And as we said, Drew, it's all yellow and green. Yeah. So all these tools, Mid Journeys, feel like Oh Ref has gotten better, but this is next level. This is like, you can take it into a scene. I don't know if you've tried some of this stuff with it yet. If you have a still, let's just say it was that girl. I tried it with something the other day where I was just like, keep this same exact image. Let's swap it to an 18 millimeter wide angle, push in closer.
Play from 0:51:54 like. The field of view is still exactly the same, but the presence of the person is bigger. And it just went right in, didn't change anything. It was like I was in a 3D sort of like space. It's like, that's crazy. Or if you do the, like I was shooting a room and it was a room, oh sort of eye level composition of the entire room. And I was like, give me an exact overhead of this. And it just flipped and the dimensions were perfect.
Play from 0:52:22 Cause I ran it through key frames. in frames and like it went from that sort of like eye level view and like traveled up to the overhead look and like nothing morphed meaning like the dimensions are perfect right so it's like that's that's what we're talking about i also have a good example of this with the car that i just shared earlier by have a preference on the video model for starting in frames?
Play from 0:52:50 I'm 2.1 now. Um, I things really good. Yeah. Um, like obviously the O V O three can, you can't, do start an M frame with V three. So I'm hoping they're going to sort that out soon. Surely they are going to, but it's, it's a bit frustrating because you can obviously do start from three and it's amazing, but yeah, like the stuff I've tested on clings pretty, pretty impressive. Yeah. Yeah. Hey, Rory, go ahead and pull your video up and I'll kind of walk through because this is a really good example here, right?
0:52:51Frames: Kling vs others; aspect-ratio flips
Play from 0:53:19 Like you can create this starting image, right? And doesn't matter where it's from. I think all of us like to go to Midjourney. But then you've got sort of, again, going back to Nano Banana here. very natural language prompts to make the update or the net new image or whatever it is, right? So I'm guessing this is what you used in terms of like fortunate beneath her.
Play from 0:53:44 Like, is this verbatim? Like what you did with this image as the image ref? Yeah. And so, you know, it's like, again, that consistent character, there's no LoRa involved. You know, I think there was, how do I get to the rest of these? I think this was a chain. Yeah. go. Oh, okay. I see. see. I see. But this is all from this one image, I guess is the point. So, you know, and this is, you know, again, very simple, you know, prompt variations.
Play from 0:54:14 Rory would be proud of me. I kind of threw mine into a Figma I was doing. I was messing around here. But like same thing, right? You get a starting image and you know, like I love the consistency with the colors, right? Like the... yeah. But let's add a backpack. Let's em change his facial expression. Let's change him smoking a cigarette. Let's change him walking back to the house. Let's remove him from the scene.
Play from 0:54:40 Let's flip the camera angle to the other side. Let's do an aerial shot, side profile. Let's add a car. Let's have them play on the phone with the car. Let's have them smoke a cigarette on the car. Let's have them hold big stacks of money on the car. But I mean, this is this type of stuff though, right? That wasn't really all that possible unless you kind of wanted to utilize chat GBT, which again kind of then takes you back into this.
Play from 0:55:09 It's not, it doesn't feel quite, yeah. It doesn't feel there. Yeah, yeah. one image storyboard, right? Like you used one image. cool. That's kind what I was diving into with that. And I think there are some limitations. think the one limitation I ran into too was aspect ratio. ah which I think I found a answer for, which was just like, if you've got, maybe like one of these is a one-one and then let's say you wanted to change it to a 16-nine.
0:55:11Aspect-ratio workaround (whitespace expand)
Play from 0:55:43 I mean, you could obviously do that in mid-jury, but if you wanted to stay in tool, ah you know, think you could act, you could add white space on either side, let's say it's 16-nine and have it fill it in, instruct it, versus saying change to a 16-nine, because I was getting, I was not getting great results with that. apparently you can, let's say you've got a two by three image and you want to, like you said, you want to make it 69. You can literally upload a 69 empty frame and it will do it.
Play from 0:56:12 Yeah. right. OK, so that's what I'm saying is like, it's just like, but if you just say it without providing that negative space, it doesn't. Like that was one limitation. I think like the other limitation was uh in Gemini, if you're doing this, uh it looks like a little bit lower res. Maybe it's a compressed file is what I was kind of seeing. don't know what's going on there because the LM Arena stuff was like insane when we were when it was testing it and then it was like a little bit like huh did they did they just release like the the fast model is that what this is versus the like the full quality model which I don't know why they hold back on that unless it's really expensive but you know it is really fast like that's another thing that we should mention here like it's fast like GPT takes like two minutes like this thing is like a couple seconds and it's
Play from 0:57:01 good but you get one, yeah, you get one image that takes minutes, like to the point where you don't even want to stay on the screen, right? That's why. I did a shot, I did a shot which was like, was, so I basically uploaded a reference of a guy and I had a, I recreated this like pickup truck and the shot that I wanted was like this guy driving like in the like driver's seat of a Porsche but you just see like the dashboard and kind of like his POV.
Play from 0:57:30 I managed to get his reflection, just the top bit of his hair in the rear view mirror and all that kind of stuff like this. It's so good. It's ridiculous. Just by saying, show the inference of his head, top of his head in the rear view mirror. Done. That's awesome. Let me pull this one up so everyone can... This is a good test in terms of consistency. To see actually how consistent it is.
0:57:47Keyframes consistency test (the car roll)
Play from 0:57:58 If you've ever run keyframes before, you know that if the two subjects are different, you'll have morphing. If the dimensions, for this specific shot, if the dimensions for the car, if there's... Let's pull an example here. If... This is if this is a different size, if this, you know, rear view mirror is in a different place, if the spoiler is a different shape, like you'll get morphing for key frames.
Play from 0:58:27 So this is like a good sort of like barometer test to show that it's like, OK, this stays the same even as it goes to an upward sort of format, right? Like everything stays the same. Even if we go down, like the wheels stay the same. Everything here stays the same, like the lines on the side of the car. stay the same, the exhaust pipe stays the same, morphed a little bit there. But again, like, this is pretty much like 95 % consistency to get something like this, which is insane to go like, all right, let's go get these angles and turn them into keyframes.
Play from 0:59:01 Because it's interesting, like with the, where is my Weavy board? I'll pull this up because this is... Rory, right? Like you're trying to shorten the gaps of these frames, right? So that there's less uh room for interpretation between those frames, correct? Correct. So it's also like keyframes. Again, if you have tools that can actually do the keyframes, which is great now, ah like that are that are actually like handle it from scene to scene.
Play from 0:59:31 The hardest part is the consistency. Like if a character's beard is two different sizes and you want to go from keyframe to keyframe, like you'll see the beard like grow in real time. So it's like you need and like keyframes is like ultimate control in terms of like start and end. But, you know, taking this and then like starting to batch process also like this is This is nano, but like on a different scale. So we're showing like the regular scale and then this is like sort of what we can do from like a batch processing system ah where this is a little tip for anyone.
Play from 1:00:02 To start with nano, it's not just the single image. This was the original image I in Midjourney. This standard image. Now, if I want to create all different angles or different shots of this car in different environments, it has the frame of reference of the front three quarter view, right? Like it has no idea what the back of the car looks like. So if you're trying to do like, you know, shots from behind or whatever, it's always going to be different because it doesn't have any frame of reference.
Play from 1:00:31 So like if you're going to try to take one subject or, you know, something like a product or a person, I think this is just like a useful sort of idea is to take that one image and then turn it into like a get like sort of the rear view of it, get the side view of it and get like a rear three quarter view of it. with, um whether that be a person, a product, or whatever. So you have the frame of reference for each side of it first before you go and iterate a million different ways.
Play from 1:00:58 if you're looking for extreme consistency, course, this is on the more professional side of things versus just the creative side. Because when we do this, I'm building little prompts that are essentially saying, take this car and just turn it to, turn this into three prompts, one for a side, one for a rear, one for a rear three quarter. Like this is all the prompt is saying with the reference image. Keep this exact car in this exact environment. Change perspective to side profile, to side angle.
Play from 1:01:24 Right? Nothing crazy. And then, you know, where it's... made like, how does it know what the back looks like? Like it's so good. It's just interpreting, And like, fits like it fits the lines. It fits the design style. Like it just, it's nails it, which is crazy. And then, you know, taking it here and then basically iterating like front shots off of this, off of this image. So using this image as the reference. And then we can get like five shots of the front of the car.
Play from 1:01:52 You're doing the same thing for the rear of the car variation, doing, you know, same thing for the uh side view. and then doing the same thing for the rear three quarter panel view. again, we're a lot of different tools, using a lot of different system prompts here to get that done. it's crazy because I'm testing it with other cars. Like we use the BMW. Works the same exact way. your core image, your kernel, then you've got these shortlisted number of pillars, right?
Play from 1:02:26 So in your case, you've got three, and then it's the big waterfall after that, because you've filled in sort of this, I don't know. let's just call it a panoramic 360 view of what this car looks like. And now, because you have identified fully what this car looks like, it can go out, in, up, down, right? um So it's almost like, I'm looking at this like there's an image up top and like this is horizontal.
Play from 1:02:54 I'm almost thinking of it like vertically. It's like the image up top. Then you've got this second layer. then you've got sort of all the iterations that are possible after that. Yeah, pyramid, yeah, there you go. Because then even more complex stuff, right? Like this isn't an easy image to play with. Like that's not an easy one. Did you do a post on that? that translucent. Yeah, you could see a lot of things in there. Amazing. So it's able to do this and it's able to sort of take that and run with it.
Play from 1:03:22 Like even just getting, you know, some of this stuff is impressive, right? Like it's getting the coils, it's getting the, the translucent, getting the gears in the front, the tire, the tire consistency, like the treads. I mean, it's so impressive with this stuff that I'm like, damn, like this one messed up here. It's the only one out of 20 that it really didn't, uh, really didn't mimic what's happening. what's the, you know, are these pretty much one-shotted for the most part?
Play from 1:03:52 Yeah. So this is, first off, you know, again, building in Batch right here. If I was to just run this workflow again, all I'm doing is dropping in this image. I'm not prompting anything. Like this is automated, right? so yeah, change the image. This will then turn into three different shots. Then it'll upscale. Then this will get routed to here to get basically changed, varied five ways, varied five ways, varied five ways.
Play from 1:04:20 It's the same exact workflow, just duplicated four times. Like this is the same exact thing. Prompt to LLM to split to different images, prompt to LLM to split to images, right? So it's not crazy, but like here in the prompting, if I wanted to, like I just gave it a shot list. That's like what I'm starting with. Just like perfect front view, macro wide angle close up, wide high, know, wide and high.
Play from 1:04:45 Same thing here for going from the side view, you know, super wide low angle. close-up Dutch tilt. So I'm giving it just a shot list of what I want. You could also automate this with an LLM if you wanted to. I was just giving it what I wanted. Then building. Yeah. I wish I had this for Jonah, because I bet you could apply that to like a shot list for characters, you know, like, you'd literally... Yeah.
Play from 1:05:10 This one should really help everyone see sort of the power of this stuff. this is where I, know, when we're talking about art creative or like creative direction for professional versus like having fun, you know, like this, need like certain control, right? So like here, this is like building out, like, you know, controlling each aspect. of the character, just like having a little section for it, the same way you might break it out into like a modular prompt.
Play from 1:05:41 But you know, hair and features, body type, age, know, ethnicity and gender, right? So we can sort of... in a little bit, Rory. There you go. Yep. so we have age, have hair and features, body type, know, style. So we can create like a base character, right? And it's like, all right, we can just sort of choose. We can run this to four different image models to see which one produces and then roll from there. You know, down here we can put essentially like a section of like clothes so we can like art direct the clothes, right?
Play from 1:06:11 Where we have this can be swapped. What this is doing is just saying, describe this clothing in detail. So we have an image. and a piece of text that sort of support it, right? And you can swap any piece of the clothes if you want to, top, mid half, lower half. And then we combine them all. We combine the images of the people and then like the prompt itself to sort of create the guy in the clothes. And once you have like the one image, right, of like the guy in the clothes, then you just go and vary it.
Play from 1:06:41 So here I was using, you know, in the same way as that car, it's just basically saying like, all right, generate. six more prompts for this guy in some sort of environment with some sort of rules. Prompt structure, basically saying structure like this every time. Run it with the LLM so it creates all the different views. Right? So I can basically do like an entire photo shoot here of this guy. Right? Like this is, and that's one way, cause we can also then throw it into, yeah.
Play from 1:07:12 You know, throw it into a storyboard format. Be like, all right, you shot list, whatever you want, character. doing it that way. Can also test like a different character with different looks to see how they want. all this stuff can happen in relatively one click. All I'm really doing here is inputting the character features and the clothes. And there you go. And then like you push it, you push go. The rest of it happens. So that's a little changing by the month over here.
Play from 1:07:37 Did this is I'm so addicted to Weavy It's insane. I gotta I have to cut myself off sometimes. Or else I'll be I'll be just totally in the rabbit hole. Just like That's so cool though, like the stuff I've seen you share on Weavy is like amazing. Like I've been tapping you up on it. I need to dive into it a lot more because I think when you have like creative ideas and when you like, there's like, like you said, there's different things.
Play from 1:08:07 like dabbling on a personal kind of just vibing with stuff versus I've got a project like Jonah, for example, using that shit is. next level. That's how you like automate stuff and get things done so much quicker. And like, I think the hardest part of everything is the setup, right? Because once you have the setup and structure, can, like you say, you drop in a different card, drop in a different person, clothes.
Play from 1:08:35 That's what it's all about. modular like that, right? Like if you can control the character, then you think about doing the same exact thing for the environment. Think about doing the same exact thing for a product. It's like, okay, and then you can composite it in there. then sort of, you know, another thing I've found relatively awesome, by the way, just for anyone who's still hanging around, you know, thinking about video stuff. You can like, with VO3 specifically, like I can cut out like that guy. like if he's just like in a position to remove the background and just like paste him in a scene without having to like relight it or restructure it or anything.
Play from 1:09:06 So it's like 2D flat guy in scene and just like type what I want to happen. And they like immediately integrates a bit like the lighting. Like so it's it's like simple things like that sometimes. It's like I don't have to go and create the whole base image with the character and the setting and everything. It's like drop it in and just be like guy walks to the right. And it's just like he automatically sort of picks it up. Um. it won't recognize that. But let me show one, this is the last thing.
Play from 1:09:32 say I need to because I really need to. This is it. Just so everyone can see, there's a structured way. And then there's how Drew and I make our thumbnails. This is the full brain on total overdrive. Let's just go in a million different directions. So everything doesn't have to be structured in here either. It can be fun. you outfits for my caricature.
Play from 1:09:58 I can put it on now. We have all our variations of our characters and our stupid text. That was all I wanted to show is that it doesn't have to be this structured workflow in there. There is the freedom to just go wherever your weird brain wants to take you in there. yeah. Well, yeah, let's wrap this. Let's wrap this baby up. Marco, man, thanks for thanks for hopping on last notice to like we enjoyed having you, man, and give this guy a follow.
Play from 1:10:27 We'll link some of the stuff to that we can. um I'll include some of those links too and send me over anything. I'll make sure to include it in the description. um yeah, dude, keep up the great work. Amazed by all the stuff that you come up with. I mean, you're one of those, I think, that always create just real high quality things with whatever tech is currently available and the quality that is available, you're always achieving at the maximum level.
Play from 1:10:54 So props to you, man. Keep up the great work and thanks for joining us. Rory, you want to say anything before we go? annoying. I'll come back whenever you want. Come back and hang. Yeah, dude. And keep creating just for no agenda creation purposes. The stuff you put out is awesome and it doesn't need to have any purpose. Best advice for anyone else here too. So Marco, appreciate it, man. Thank you for coming.
Play from 1:11:20 For anybody who's still hanging on, if you haven't freaking liked and subscribed by now and told your mother and told your dog and told your boyfriend or girlfriend or whatever, just go do it. Help us out. Come on. Another episode where we waited till the very last second to throw that out there. Also, next week is gonna be episode 50, so we're thinking about doing a live stream for that. Come join us and come hang out. And we'll catch you guys in the next one. Later.