Ep. 73: She Cracked AI Advertising Before Brands Were Ready

0:00:00Meet Salma Aboukar

Play from 0:00:01 8:88:88 0:00:01 What's up, AI Nerds? Fast Hours Podcast, episode 73, Rory Flynn, Drew Brucker, and we have a special guest today. Uh I would even go as far as to say Twitter X legend. Salma Aboukarr is with us today. Creative director and founder. We're excited that she's here with us today. I think this has been a little bit of a planning for us to get you on here.

Play from 0:00:27 8:88:88 0:00:27 Super excited to get your insights. um across the board. And I think anybody listening to this is gonna get a ton of value out of it. So we're really excited for this one. But Salma, why don't you do us a favor? Uh give us a quick intro about uh who you are, what you're doing, uh so the listeners can get a little bit of perspective for today. Hello, everyone. I'm so excited to be here. Um I'm Sam Abica.

Play from 0:00:53 8:88:88 0:00:53 I am the creative director and founder of Creates. Um basically working with brands mainly on projects, so whether it's ads, video ads, things like that. I've been doing this for the past two, two and a half, three years now. Um worked with like major brands like Coca-Cola, Svetka, um and also like Google Labs as well and Panasonic. So really, really excited to kind of like talk about AI and yeah.

Play from 0:01:21 8:88:88 0:01:21 Isn't it isn't it amazing the brands that you've been able to work with just from putting yourself out there? I think that's collectively in, you know, everyone here. Um, we've worked with some clients that I think in some logos were just like, wow, you know, like looking back on it, that's pretty awesome considering we were just sort of being curious and and being creative and we were able to leverage it, you know, in a different way.

Play from 0:01:47 8:88:88 0:01:47 And I would love to catch up too on what were you doing before? This, like what is your background? Rory, that's I feel like that's typically a Rory question, but I'm gonna steal Rory's question for this. Give us, give us a little bit of background as to what you were doing prior to AI and and maybe even like your inflection moment, your Rubicon moment that that sort of changed things for you in terms of maybe your pivot in your career or sort of like the blue ocean of of what's possible for you.

0:02:17From CGI agency to AI-first studio

Play from 0:02:18 8:88:88 0:02:18 Yeah, so I was actually running crates as an agency. So we were doing uh projects with brands. using all CGI, so Blender, 3ds Max, creating these beautiful environment scenes and putting products in there. So it was mainly like an agency which was driven uh using you know CGI at the time. And what was interesting was when Midjourney first came out with their like really terrible model, I think it was version three, um I realized you know maybe we can start using AI to generate at least the backgrounds and the backdrops.

Play from 0:02:50 8:88:88 0:02:50 And that's when I really started taking AI seriously and started like Like experimenting and putting out my findings on X. Um, and really that's where it started. Like, I thought, okay, if AI gets better. Then it's gonna really change how we do things. And to the point where we started actually building our own tools around AI to better execute on these projects. Um, I was still using 3D right up until I'd say last year.

Play from 0:03:16 8:88:88 0:03:16 Um, and then when you know the video models got really good, um, then like right now we use 100% AI for everything. So that was the moment like when the image models got better. We started doing workflows with stable fusion, um, using backdrops from Midjourney, putting products in with 3D and then once the videos got better we kind of like pivoted to start doing video ads as well. That was definitely one of the first things that caught my eye when I I think probably when I just joined X.

Play from 0:03:45 8:88:88 0:03:45 You're one of the only people I saw at the time that was doing direct to consumer stuff. And that's, you know, my background also from the e comm space. And I was like, this is this is cool. This is right up my alley. And uh then it started digging in and saw you, you know, you were building your own product at the time. And it looked like I was like, this like whoever this is, they get it. Because they like they understood that it was like, we need to put a product in a location.

Play from 0:04:12 8:88:88 0:04:12 And then we need to do that across like all these different types of, you know, uh like basically a whole photo shoot within you were one of the first people that I saw and you were trying to at that time it looked like you were blending 3D and AI together. And that was early. Like it's not as easy as it is now. Like that was you know, it's like how are you doing that? I remember just looking at it being like, this is directionally exactly where this needs to go. But like how do like from that that project, I'm just I'm just curious.

Play from 0:04:38 8:88:88 0:04:38 I'm sure people would love to know like back then a little bit. What it w like the the challenges, the struggles around building something like that. Yeah, I mean so what really inspired it was like I've been working with brands for a few years at the time and one thing with AI is lots of DTC brands, you know, don't have the appetite for things hallucinating to the labels and the logos. Um so I thought, okay, well why don't we kinda have a hybrid model?

0:04:44Product fidelity before AI got good

Play from 0:05:03 8:88:88 0:05:03 So we were actually creating 3D models by hand at the time and then imported them into a AI scene and then using um clever techniques to match the shadows and everything else. But the idea was to ha you know keep the product fidelity, the logo and everything else. um So that's like how we kinda approached AI, kind Of you know, in hindsight, it made sense, but also it was a scalable because we were doing uh things in a you know in an interesting way with 3D, and 3D took a really long time.

Play from 0:05:33 8:88:88 0:05:33 Whereas now, um, now that we have like different models, 3D, like rodent models that get like the actual products right, there's still issues with the labels, so we still kind of have our own our own internal tools to kind of fix the labels and last 20%. 'Cause was I remember looking at the that was like the it was like almost like the model for what Photoshop eventually did, which was like, let's create like here's item, turn item to three D, like readjust lighting.

Play from 0:06:01 8:88:88 0:06:01 Like that was sort of like what Photoshop tried to do. I don't know if that really took off because that product was kind of meh. But like I like I r saw the vision and the vision was correct. I think that's where a lot of uh you know, I think it was just like everyone was striving for this. product stuff and then it didn't come along until you know we could really have I think personally I think it was the higher resolution that we needed for smaller text because it just didn't have enough pixels to work with.

Play from 0:06:28 8:88:88 0:06:28 But I was always I remember like it was I immediately contacted you on X. I was like what what is what is this? I I need to know what this is. But you've also been a you've also been a Midjourney OG. Like back probably one of the first people I think I've found on X. There wasn't you know, at the time it was a little bit different. But of it so you were using it at you were using it version three or you started really like pumping at version four?

Play from 0:06:55 8:88:88 0:06:55 started pumping it version four and then version five. Uh version three was like kind of like testing to see, you know, if this is even a thing. but yeah, it was really version four and five. Um and I would generate the backdrops first and then kinda like uh add products, style it and then render it. They were I remember uh you were doing a lot I feel like it was like skincare. That was what caught my eye.

Play from 0:07:20 8:88:88 0:07:20 'cause you were doing really great work early on with like skin texture too, and like not just focusing on like here's AI model. It was like it felt like it was conscious of how the skin should look, eyebrows, things like that. Yeah. What's so interesting then was I remember um stability, you know, they had that model, stable diffusion, and you had to download it locally with automatic one or was it one one one automatic um thing that you download?

0:07:34The duct-taped road to photorealism

Play from 0:07:47 8:88:88 0:07:47 And um and creating a whole workflow to kind of get the skin texture, the eyebrows and everything look incredible. And this was shortly before mid um Magnetic came out with their upscaler. Um but that was one of the the the tricks that I was using for my clients to kind of get everything to look really real and the workflow was generate images on Midjourney. Midjourney at the time had like very like plastic skin. So what I would do is upscale it with stable diffusion, have a workflow and then kind of like colour correct, adjust make it look very realistic.

Play from 0:08:19 8:88:88 0:08:19 But now, you know, with especially with the latest model on Midjourney, um you can get the skin texture and the hair texture and everything beautiful just by Pompton. Really liking the skin texture right now. We're getting there. I mean, it's it's pretty damn good. I mean, isn't it funny though, looking back on it, all the workarounds that we had to duct tape together to get things to the desired outcome that we were looking for at the time?

Play from 0:08:45 8:88:88 0:08:45 The one the one example still cracks me up. Rory, remember when you were doing the the character consistency with multiple characters inside of Midjourney and this was like V5? And there was this whole workaround that, you know, was just so unnecessary to get to that point, but you had like that you had to do that to to get anywhere close to it. And you did a whole breakdown on it and it was just like, my God, like this is awesome, but this sucks.

Play from 0:09:14 8:88:88 0:09:14 This sucks that we have to do this. Just wait six months, it'll be out. You know, s so some someone could do this. Do you so do you have a background in skincare or like cosmetics or anything like that? Okay. uh as an agency, I mean we used to do s we didn't do like actual like model faces but we used to do hands hand textures and create 3D out of that and it was I mean if you know the workaround with AI and how you duct tape it like the 3D version was way worse like it would take me weeks just to kind of get a hand to look somewhat human.

Play from 0:09:51 8:88:88 0:09:51 So yeah I worked with a lot of like skincare, cosmetics, uh CPG drink brands, sweet brands and packaging mainly. Um and it was more like you know for their website sites, social media, ads, things like that. So when AI like came around, it was, you know, it was I was on it. Like I was like, okay, if we can kind of like get rid of the 3D component, that would be awesome. And you know now we can.

Play from 0:10:17 8:88:88 0:10:17 We don't need to use 3D as much and there's often 'Cause I I would say because I the reason I ask is 'cause I look at a lot of times like the composition and the art direction of what you're doing, and I'm like, this this mimics I feel like a photography brief from what you'd get from a cosmetics brand. Like this like that type of shot, that type of cropping, framing, the angle, like the the lighting, things like that. It felt like was very technically sound, not just like skincare ad, enter.

Play from 0:10:47 8:88:88 0:10:47 You know, it felt like there was a lot of detail behind it. Yeah, yeah. and that's what drove me like to like d there's more behind the scenes. You know, like there's definitely more behind the scenes and there's more going on 'cause I think that was probably like how did you how did you like sort of come across the control aspect of Midjourney back then? Like what were you doing to to sort of like control your vision? a lot of reference images, a lot of like describe feature, um, also had like a prompt system going, um, where it was I think it was something like subject, detail, um, composition, background, lighting.

0:11:10Art direction beyond basic prompting

Play from 0:11:24 8:88:88 0:11:24 Um, so it was a lot of that and a lot of experimenting. And then also sometimes what I would do is I would use other local models to kind of generate um some of the models and then kind of like bring it into Midjourney and then improve it on Midjourney and then bring it back into like stability and then improve the skin. So there's a lot of back and forth of different models. but to be fair, like when it comes to the art direction and just the creative feel, I feel like Midjourney has a level of aesthetic that you can't really get from other models.

Play from 0:11:54 8:88:88 0:11:54 Even like now, like with Nano Banana and stuff like I I'm not sure what yeah, like you know what I mean? Like it's the models kind of look the same. Um unless you like distinctively ask for certain features and one of the tricks I actually use sometimes right now not as much but before before Midjourney version this latest version I think is the eight point two what I would do is I would actually kinda like use Chad GPT, uh bring in an image of a model with like very distinctive features.

Play from 0:12:26 8:88:88 0:12:26 So like maybe a gap T, you know, uh certain look and sort of signature pieces that aren't just everyday things. Exactly, exactly right. And then and then what I would do is I'd then I would take a Tanana banana and recreate that image. And then a little trick I'd use is I'd add like a little bit of a like a maybe a tattoo or an imperfection to make it look, you know, real or like stylish.

Play from 0:12:54 8:88:88 0:12:54 Look at the skin on that one. Like that one to me again, like just there's there's thought behind this stuff. Like I feel I I Yeah. just love even the little ring here, right? Like that's so real. Yeah, I think with that one I actually used Nano Banana and again like the work before that was I brainstormed it with fraud um and again like picking up on the uh particles so this was really a particle type shot.

Play from 0:13:22 8:88:88 0:13:22 So it's picking up on the particles, how would the particles look like on the skin? So the little details that come in on the skin and then now here like as well the little particles, these are the little things that I kind of like to capture. Those are the things that make the great images different from the good ones. There's the intentionality of that extra layer, right? Like the fact that you thought about that, um, and it's visibly distinct here.

Play from 0:13:53 8:88:88 0:13:53 It's not maybe the main focal point. It's a supporting element that if you're looking at this longer than a second or two, you can actually appreciate it even more. Right. And I think this is really cool. And I I do notice that with your stuff in general too. I think that's a really good call out. Are you um maybe give us a a rundown on the tools that are currently in your stack right now? I feel like everybody always wants to kind of see what people are using and maybe even the why or how you think about using each one.

Play from 0:14:23 8:88:88 0:14:23 Like does Midjourney serve a certain purpose for you versus something else? Right. Yeah. No, for sure. I th right now I have uh a bunch of skills on Claude. So what I like to do is create skills on certain things, skin texture, uh models with distinct features. Um I also do it f a lot for videos as well. Um and then Midjourney is purely used as like a creative stylist. It's just for aesthetics, finding new styles, um, and then the other models are I use it as an editing.

0:14:28Salma’s current AI creative stack

Play from 0:14:54 8:88:88 0:14:54 So Nano Banana as an editing model, and then um so if I start up from the top. what I would do is I'll brief with uh Claude using my skills, then I would get my prompts and then I'd throw it into uh Midjourney and then download it, edit it on Nano Banana, and then kind of like clean it up, upscale if needed as well. It it's a wonderful one. The the thing I love about with Claude that everyone has like everyone's got like these like little micro skills that they use for like these these little things they do.

Play from 0:15:26 8:88:88 0:15:26 But like that that probably makes the whole difference is like having a skin texture skill. Like that makes like like all the difference. You know what I mean? Like people aren't taking it that far and and everyone who's listening, this is what the is what a professional sounds like. This is the level of detail that you have to go into. on certain things to like make this stand out from just like skincare ad enter. Um, you know, skills for s you know, skin care for skin textures, skills for uh, you know, model selection and unique features.

Play from 0:15:53 8:88:88 0:15:53 Like it's probably the best way. Do you find it when you're doing this, do you find it better to have like these isolated skills, like in these certain little specializations, or just like have like model generation as like a as like a skill itself? Like it are I there's no right or wrong answer to me. I'm just curious how you how you're finding it. I I actually separate them just so that I can keep track. And then I also the other thing that I've been experimenting with recently, which would be valuable, is you know, with so many models coming out, you know, how do you easily test the model?

0:16:08How she stress-tests every new model

Play from 0:16:23 8:88:88 0:16:23 So I created another like skill that kind of evaluates each model. Um so that skill basically, whenever there's a new model, we just run it through all the models and see like which areas is weak, which areas you know improved on. And that I use it mainly for the videos because as you can imagine, um Seedance is amazing, amazing model. But there are some areas like VO3 is still kind of good at, ah but for the most part I would say C dance is like really, really, really good models.

Play from 0:16:52 8:88:88 0:16:52 So I I have that skill to just kind of evaluate any any new models that come out, any new image models or video models as well. Is that something you're you're running through like an C P or you're using that in like foul or are you just doing it manually? I am I run it through foul. Yeah. one. If anyone if anyone's not familiar with with foul or what it is, it's just like an API aggregator. But if you have Claude, you can tap into the MCP and basically just tell Claude to run this in C dance and it'll hit the API, come back and run it directly in your in your chat.

Play from 0:17:26 8:88:88 0:17:26 So if you're looking for something that wants that you pay per play, you don't have to pay for a subscription to it. You just pay per API hit. It's a great option for anyone who's looking out there. Just gonna caveat. with that. But the um the other thing I wanted to definitely wanted to ask you about was on the video side too. Like there's to me, I hate the video upscalers.

Play from 0:17:52 8:88:88 0:17:52 Like I they take everything that's good up to me about a video and make it worse. Like a all of them smooth everything out for someone who's doing skin texture, I'm sure this has to like piss you off a little bit. Like the Yeah. Yeah. Um one thing I do is I actually generate the videos uh sometimes in four K but other time just uh ten eighty.

Play from 0:18:17 8:88:88 0:18:17 Um and then colour correct the keys the all in the colour correction and the contrast. and I find that works way better than upscaling videos uh because otherwise it just looks so artificial. It's i it's really annoying that that's the the case. It's like, we can add more definition in everything but skin. It's like it seems like it adds definition to just about everything else, but it does make it the edges become so obviously AI.

Play from 0:18:45 8:88:88 0:18:45 Um d like you said, too sharp, too much contrast. And then like for someone who's doing skincare, I'm sure this is it's just like that's to me that would drive me nuts. That would I would go absolutely insane. But are you are you finding any other any of them? Have you tried any of the other ones, like the open source upscalers for video? Like I've I forget what the the other one was. People reason through comfy a lot. L LTX? I think LTX is doing some incredible stuff as well.

Play from 0:19:12 8:88:88 0:19:12 I I avoid like I try to avoid any video upscalers for those reasons. I I mean we just use Premiere Pro to just like colour correct and adjust things. I find that to work really, really well. And then clever editing. I think also the editing, you know, could does a little bit of the heavy lifting when it comes to that. Um so yeah, like I don't currently I don't use any um upscalers, but before C Dance and before these models and I when I was working on another campaign late last year, um we had to create like these workloads.

Play from 0:19:45 8:88:88 0:19:45 around certain things um which wasn't skin related it was more related to like animals texture things like that and it was a little bit easier to work with um But you know, again, like we did upscale a few of the like so with what one of the ads that I worked on, which was the Coca Cola ad, we had to upscale the trucks and you know, as you can imagine, it adds a level of sharpness, but then we edited it and adjusted the colours to make it look uh nice and more photorealistic.

Play from 0:20:13 8:88:88 0:20:13 But I do feel like n with the these newer models the output would probably be much stronger. And this is also a reminder for everyone color correct your AI videos. Like just do something with it because you see it even just the slightest adjustments make a big difference. Yeah. that's another one of those things that's super available and accessible that can really make a big difference in a big way.

0:20:20Why color correction still matters

Play from 0:20:41 8:88:88 0:20:41 I've been thinking about that a lot more too, like going I I used to do that a lot from the still side. But you know, now some of the color grading that's baked in has gotten so good. Maybe you don't have to do it all the time, but um, there's still an opportunity there. Like I almost think about it from the sense of Depending on the type of project or work that you're working on, almost the photographer approach of post-edit uh of a photo, right?

Play from 0:21:07 8:88:88 0:21:07 Like thinking about, I don't, I don't know how familiar you guys are with Lightroom, but you know, thinking about adding linear gradients in certain spots, dodging and burning, you know, uh luminance adjustments, right? To just refine the image. I mean, there's an opportunity there to even layer on and polish images up. You know, versus just taking exactly what you got out of the model and then rolling with it, right? Like I think it's that's sort of what separates I think this idea of, anybody can create anything and versus like this artistic mindset of, okay, now I've developed the raw material, like that core raw material of what I'm gonna use, and now I'm going to layer on top of it.

Play from 0:21:47 8:88:88 0:21:47 And that doesn't have to be just stills, right? But I think it's like that mindset of finding ways to tweak it even further, to make it even a little bit more yours, right? And differentiate it from the space because, you know, granted, assuming everybody in this room is a great prompter, right? What's stopping you guys from creating the exact same thing as me? Just the idea, the concept that I landed on. But assuming you saw that and you wanted to use it, you could generate the same thing.

Play from 0:22:14 8:88:88 0:22:14 So I think like those additional layers and editing and adjustments, um, even the application of it, um, can kind of set the stage for the final outcome. But That's super interesting. I also want to touch base on, you know, Salma, you had like at least two really big moments that I know of, probably more, um, on X. I mean, you've been on there a lot. I think you got like 50k followers. So, I mean, you you've been doing your thing regardless.

0:22:25The IKEA video that changed everything

Play from 0:22:42 8:88:88 0:22:42 but a couple things that stand out to me. You probably, if you're listening to this, you probably almost guaranteed, actually, have seen the whole IKEA explosion video that went. Because that was I you I mean you got picked up I think nationally for stuff like that, right? Like there were tell tell us about that. I would love to hear this is gonna be a little bit of a loaded question, so I want you to take your time with this, but I'm thinking like take take us through one, how you came up with that concept.

Play from 0:23:14 8:88:88 0:23:14 Because before that, I don't know if I saw that concept in the video medium. at that point. Like I it was it was a novel thing that I saw, number one. So I would love like that concept ideation or what gave you inspiration to come up with that. And then two, I would love to like hear your perspective as something went so viral, what were the good parts about that and what were the bad parts about it?

Play from 0:23:42 8:88:88 0:23:42 Yeah, I mean, so was I think it was last year summer, um, and I was seeing like a lot of people posting about VO3. I think sorry, V VO is it VO2, yeah, VO3. Um so VO3 just came out and you know, I was seeing all these posts and I was thinking, you know what would be interesting is to do something that is, you know, some new idea. And I was just like researching and checking out some stuff and I saw a like a box.

Play from 0:24:08 8:88:88 0:24:08 I think it was an Amazon box that was flown in the air and it was transforming into like uh like a shoe and then it transformed into like a top, something like that. So th so then I thought, you know, why don't I kind of like do something for furniture. Um, what if we have like a IKEA box in the middle of a room and then it kinda expand? So I started prompting and, you know, the first few videos I got were really bad. Um they just look nothing like what you're seeing here.

Play from 0:24:34 8:88:88 0:24:34 And then I was thinking, you know, can we push this model like far? Like how far can we push this model? So then I started using like really detailed prompts to kinda get it. And it was it would get some things, but it wouldn't get everything. And then I I thought you know what? Well I saw something about JSON um but I couldn't understand what it was at the time. So then I just gave it like my I think the LM I was using was probably ChatGPT. I gave it my prompt and I just said, you know, convert this detail props into JSON.

Play from 0:25:03 8:88:88 0:25:03 I want to make sure I have control over, you know, how the box explode, it needs to be seamless. Uh we need to make sure like certain elements like for example the throw comes in last. So kind of like had timestamps within it as well. And then and then it came up with this long JSON prompt. And then I threw it in there and honestly came up with such an amazing video. And I thought let me post it, let me just you know, put the prompt bearers up and just you know, didn't really think too much about it.

Play from 0:25:34 8:88:88 0:25:34 And then immediately everyone started doing the exploding boxes. Um I think it got yeah. everyone. Me, I think me too. I was like, this is cool. Like what what can I do with this? Yeah, this is awesome. Yeah, it was amazing to see. And then I think like a few people wrote like articles about that. And I think it was for the first time ever, maybe, when it comes to videos. Um, it can be used for brands, right?

Play from 0:26:01 8:88:88 0:26:01 Like, okay, you know, this is the video that I think made brands feel like, okay, v video has some serious leg. Um, and then I got contacted by a bunch of people, got contacted by Silbicide to work on um, you know, the Panasonic campaign, the Coca Cola and then also the sequel. And did did IKEA reach out to you or did they just say, you know?

Play from 0:26:26 8:88:88 0:26:26 They did. Good. to me and and interestingly enough, they reached out to me. And I think at the time I was getting so many messages that by that time I already was signed on to another company of the two. Yeah. Right. Yeah. Um and what else? And then the downside, I mean, you know, I think The the I think the only downside that I experimented as well, experienced the at that time was, you know, I don't know if there was much negativity, but it was more so, you know, just the AI doomers that were saying like, these are not like this is not the exact furniture or like this isn't exactly something that is a hundred percent usable.

Play from 0:27:09 8:88:88 0:27:09 Um, but apart from that it was all round positive. Th I always have I always hated this like usable situation where it's just like, Yeah, well, the unusable one got millions of impressions and the usable one that will look distilled from this, you know, a thousand times and look like crap because it's all on brand. Like five hundred impressions total.

0:27:13Going viral and handling AI backlash

Play from 0:27:34 8:88:88 0:27:34 You know what I mean? It's just like sometimes you get you gotta like you gotta stir the pot a little bit. But do you get do you get much hate still on X? I feel like it's Sort of died down. Yeah, it it has. I mean, I've been a little bit strategic by not posting some things that all, you know, completely off. I I reserve those for LinkedIn, you know, LinkedIn everyone's happy.

Play from 0:27:59 8:88:88 0:27:59 There you go. Surprisingly I get more on LinkedIn now than I do on X. Yeah. least on LinkedIn, they gotta uh they gotta have their little name and profile picture attached to it. So you there's some there's some built-in accountability and reputation there. You know, you can't just say any I mean some people do say anything, but you know, and and you also know it's probably a real person, you know. Some I I don't you know, I don't I don't know how many of these haters are real people or not.

Play from 0:28:29 8:88:88 0:28:29 I have no idea. And everybody's hiding behind. It seems like anybody that's commenting on that is typically not someone that has their face in their profile picture. And they've typically got less than, you know, a hundred followers. Um, just largely speaking. So it, you know, takes everything into consideration. Like, what are you hiding behind? So, yeah. like a I got a I got a comment on my my Figma presentation the other day.

Play from 0:28:56 8:88:88 0:28:56 Um, you know, the one that that they put up online and it was about my hair and I was like, that's a new one because I didn't have a hat on and they were just they were just like like really bold going up there with that hair. I was like, what the f the hell, man? Like I haven't gotten this before, you know? My my looks haven't been trashed. And I was like, this doesn't this is weird. It's normally like you're an idiot, you know, you're you're just, you know, you're killing the environment. You suck at best, you're not a real insert X Mm.

Play from 0:29:23 8:88:88 0:29:23 you know, whatever. But the first time and I was like, huh, that's why I wear a hat every time. Now no, this was on YouTube, but I know YouTube can get a little bit crazy too. Um LinkedIn LinkedIn's, you know, sometimes people try to get more technical and they try to, you know, r nitpick. Like at point three seconds, this is what happened. It's like, okay, all right. Ninety nine percent of people aren't gonna see that, but you know, I knew it was there because I tried to edit it out and just couldn't get it out.

Play from 0:29:51 8:88:88 0:29:51 But Glad you found it. so I think some of that is maybe like a insecurity ego thing. They're they're like they almost wanna find something to that they perceive as adding value in their mind, like, you didn't talk about this or you didn't notice this, but I noticed that. And it's like I don't know, that's some weird thing that's going on. But it's I mean it's even more like I would say aggressive in the advertising community because you know, so many anti AI traditional people on on LinkedIn who have comments.

Play from 0:30:25 8:88:88 0:30:25 Um but I mean I'm seeing now everyone is like adopting because you kinda have to at this point. Um but it is super interesting though to see like all the different platforms what type of hate people get. Yeah. I hey, I have a I have a random question for you. Um we were talking a lot about taste is a moat, you know, we were kind of going back and forth on what that really means, but I think the idea of moats is interesting.

0:30:38Originality as the next creative moat

Play from 0:30:57 8:88:88 0:30:57 Uh also just maybe this idea of like things that aren't tapped into yet or that creatives aren't thinking about. Are there any like of those Blue oceans, I guess, for creatives that maybe you see or would predict that would be opportunistic for people in this space, you know, and and I think that's a little bit general, but like thinking about this from like, you know, whether you're a designer, creative director, art director, photographer, anything creative, like there are any areas that you think are extremely undervalued, overlooked, underleveraged right now that you've seen?

Play from 0:31:38 8:88:88 0:31:38 Um, I would say like so originality is is an interesting uh thing that I've been thinking about, um, because a lot of the stuff, especially with AI, you know, it's kinda like just the same old thing recycled. Um and I think like if you come from a traditional background, um whether it's photography, creative direction, anything like that, that that is I can see that becoming very valuable because if you pair that up, your thinking, your style, with AI, that I think that is the opportunity to come up with something original.

Play from 0:32:10 8:88:88 0:32:10 And I do feel like right now, um, in general, I think people have seen it all. Like, you know, when it comes to AI, it's you know, before you'd put out a video or an image and the skin would look great and everyone's like wow. But right now, like, you know, they've seen it all. So I think originality and coming up with new ideas. I think that's gonna be the biggest, you know, I I guess we can call it moat, um, original concept. And also the other one is having a unique style that is kind of like strong identifiable.

Play from 0:32:41 8:88:88 0:32:41 Because with AI, you know, people, as you can imagine, I'm sure you guys content ha content has been ripped off many a times. But if it's distinct, now, you know, people are able to like so I think like originality and then actually being intentional with your style. I think those are the two areas I see great opportunities because even with brands they want to work with someone who has a unique style, who has a unique flair. I wanna can can we dig into that really Rory?

0:33:05Can AI actually be original?

Play from 0:33:08 8:88:88 0:33:08 I'll let you go. I just want to get into this originality thing really quick. Cause I was at an event last night. It was a it was an AI event here in Atlanta. I mean, they're few and far between. Um, but I was glad to be at this one. I was speaking at this one. There was a interesting question or topic that came up. And I it was, I think I actually maybe posed it, but it was like originality, like, is AI capable of originality?

Play from 0:33:34 8:88:88 0:33:34 And I think when you first hear that question, a lot of people think no. And I think my hot take is I think it can be original, but it's based on the idea that you just mentioned, which is if you pour your individual unique experience, context, expertise, right, view into an LLM at a deep level, it absolutely can be original.

Play from 0:34:01 8:88:88 0:34:01 Right, because you are original. You're pouring pieces of it. And then it can build, it can take those original, originality, raw materials, if you will, to build new ideas or adjacent ideas. To me, that's being creative and that's being original, but it's because of the person on the other side of that. So AI alone, I don't know that that can be original, but I think AI in the use of anyone, it can be.

Play from 0:34:29 8:88:88 0:34:29 if you've got the right amount of context poured into it. Would you guys agree with that? And do you have anything else that would go would support either direction? I'm curious of both like your guys' view on that specifically. Yeah, I I mean I I a hundred percent agree with that. Like I think AI only can do derivative things, right? But you as the original, like you as the creator who comes from, you know, as you said, experience, context, background, if you put pour that into an LM model and really like take your time with it, you you'll be able to come up with something new.

Play from 0:35:04 8:88:88 0:35:04 Um, but yeah, I would a hundred percent agree with that statement, a hundred percent. What do you think, Rory? I'd say so. I mean like look, the some of the things I come up with on I I do on nano, I'm like that's been done before. You know, but like things I see on, you know, not just a glaze Midjourney on this one, but it's like when you're bashing mood boards together and S ref codes and just like here's all these random ideas that never no one ever thought about before, all this data going in and then it spits something out.

Play from 0:35:31 8:88:88 0:35:31 There's I know there's stuff in my library that has never been explored before. Right? Like you know what I mean? Like this stuff in the that's Like who thinks of this? No one would think of this. It's so absurd. It's so so different. So I'm not saying I'm special. I'm just saying like I've mashed a bunch of like crazy ideas together and hit enter. but I was also because I'm like, well, what happens if it looks like this with this type of mood board? Black and white photo with a 2D vector, you know, image as a st as a style reference, and then like, you know, black and white, but then also like vivid color, but then also, you know, it's like like I just want to see what what it can do with all of the overloading of data.

Play from 0:36:08 8:88:88 0:36:08 that I'm giving it. Um so I don't you know, I think a lot of stuff is we're not seeing a lot of original like in the marketplace. I think when it when you see original, you recognize it because it's so good. I think everyone's just trying to do what we've already done, but do it with AI. It's not like we haven't explored the new medium yet to its fullest. Um, if that makes sense. But side you brought up a good point, Rory, which is like mid and and maybe Midjourney is the best at that because you've got personalization, which is a huge individual wedge itself, right?

Play from 0:36:42 8:88:88 0:36:42 There's a big difference in me using Midjourney with my personalization versus Sama versus Roy. We're gonna get three different things right there already. Right. Those things are already tailored. Now do th do those necessarily make the image unique? No, not maybe not alone, right? But it does lean into more of the things that we like. And the things that we like, there's a variable there, you know, a spectrum of, you know, do we like what everybody else likes, or do we like things that are very unique? So you've got that, but then to your point, you've got these other really variable levers like references, blending, mood boards, right?

Play from 0:37:20 8:88:88 0:37:20 that I think like 10x that originality, you know, because There are so many style codes. You we could just talk about style references alone, right? Like there are billions in Midjourney alone. You couple that you you start stacking those, those are unique. Like those are all basically one of ones. So like from the visual asset thing, I I do think Midjourney's done like a very good job of that, intentionally or otherwise, compared to others.

Play from 0:37:49 8:88:88 0:37:49 And I think that's one of those cool things that us creatives really love about that platform still. Miss it. I miss just bashing stuff together in there for no reason. Just like it's like my favorite it's my favorite tool to just like I have n I have no creativity right now. Like it'll pull it out of me. You know, if I'm just like I've been been doing writing system prompts all day. It's like, all right, let go in here.

Play from 0:38:14 8:88:88 0:38:14 And it's like but Matt, you know, with the is really intuitively great at that. Like pulling, like making you feel like you're having fun, you're experimenting, and then quickly you get something that is just a slight adjacent maybe to what you were expecting. And then sometimes you go off on that tangent, right? Or, you know, again, there's no pressure. You you can sort there's so many variables that take place.

Play from 0:38:41 8:88:88 0:38:41 It really is um A nice thing for creatives, especially when you're in the ideation phase. Like you don't even know maybe exactly what you want, but maybe you have an idea of what. Exactly. I wanna I wanna shift fr I wanna take this shift now because I think this is also interesting, right? Of going from this mindset to now like I'm sure Osama's you're running is it is it technically an agency? Is it ta your business technically an agency or is it a little bit hybrid?

0:38:52Inside an AI-native creative agency

Play from 0:39:09 8:88:88 0:39:09 it's a little bit hybrid. So we have like the i the tools. We kept it internal, so we used our own tools to kinda deliver uh the uh the deliverables. Um but at some point we're probably gonna like open it up to everyone. that's sort of like I'm interested in the infrastructure, like sort of how you have like sort of a thought about like constructing this to develop like a new age type of agency. Cause you know, I come from the agency world, ran an agency, was like a lot of people and a lot of different things that we did manually that now I would think about differently.

Play from 0:39:39 8:88:88 0:39:39 But like how are you like are are you so you're building internal tools? Are you are you doing a lot of things via via automation? Is it still pretty manual? Are you are are like people building their own stuff? that you know, that are working with you. Like I'm I'm curious to what the infrastructure looks like with the AI tools we have. Yeah, so it's like internal tools um for anyone who has access to it. Um we have a lot of different models, so like and some of it's automated. So the way I I've we've approached it is like we have aesthetic models, so obviously we don't have that mid-journey in there because they don't have an API, but we use things like Z image Turbo, which in my opinion is Saton because that's actually a very good that's actually a very good aesthetic model.

0:39:55Z-Image Turbo and aesthetic base models

Play from 0:40:19 8:88:88 0:40:19 Mm-hmm. Yeah. hasn't tested that one. Yeah. us more about that. I d I don't I've I've heard the name, I don't know anything beyond the name. Yeah, I think I I may have done a post on it on my uh X but basically it's it's a open I think it's open source but it's one of the local models. Um and what it does is it kind of adds the aesthetic feel. So the similar to Midjourney really.

Play from 0:40:46 8:88:88 0:40:46 So it has like the style, the aesthetic. The only difference is that you obviously you kind of need to um the prompt needs to be heavy, like um it is quite prompt hungry but we kind of have it like we have a system prompt for that. Um but these are the kind of images that you get from there like very aesthetic, very cool. uh Again like we're trying to simulate similar to mid-journey and then use that as a base and then edit it.

Play from 0:41:11 8:88:88 0:41:11 So that's how we kind of like approach the models. We kind of have them um kind of like put them in a way that if you're like a customer um or if well if I'm working on project I would always start with my aesthetic base models. Then use my editing models. uh to ask you about that. So I was gonna ask about that stack. So Z image turbo, you're treating like Midjourney in a sense, base image.

Play from 0:41:36 8:88:88 0:41:36 We get what we want, but if I wanna tweak something, I'm then taking it over to edit model. So maybe that's color, texture, lighting, something like that. Okay. Found it as great color. Does that is that am I crazy on that or is that something that you've noticed also? Yeah. It also has great composition. Also understands styling. Um so if you go a little bit up, there's a model, I think sh yeah like styling um when it comes to fashion it does really well as well so and then also their models they don't look like you know you can get diverse unique looks when it comes to the models as well um but overall is the just the feel and using that as a base image we find like does an incredible job it just kind of makes things feel more elevated

Play from 0:42:23 8:88:88 0:42:23 And that's that's bit that's a lot of the sort of like the ground floor of the projects, is using that. Okay. Interesting. And are you doing like go ahead, sorry. Especially like for cosmetics and any what like anything that requires styling. So cosmetics, fashion, things like that. Um z Z image Z image turbo is our go to. And and are you like let's just let's just I'm not trying to again get your trade secrets here, but just like w someone someone comes in with a brief.

Play from 0:42:54 8:88:88 0:42:54 Are you running like similar processes every time where it's like, Okay, this is how we produce this, this is how we deliver this first pass to them and then like you're building the system around that or is it kind of like project comes in, we have to build unique system for for certain projects and then we reuse them. Uh I'm just Yeah, we have an agent, so the agent takes the brief. So you upload your brief. So you know, could we upload the brief? to the agent and the agent kinda like does the initial fast past of it so understands the the brand what we're trying to achieve and then it starts to generate the first batch um using the aesthetic model sometimes it skips the aesthetic model if it's like CPG and but for fashion and skincare it goes through that pipeline and then it goes through another pipeline of editing so this will be like adjusting you know clothes maybe getting rid of certain things artifacts and then finally it renders it

0:43:06From client brief to automated pipeline

Play from 0:43:45 8:88:88 0:43:45 That's like a bad one. Yeah. me this is what makes me happy. I like that you divided it too by use case, like the maybe the type of thing you're generating too. Maybe it doesn't need a mid-journey or Z image turbo. Maybe it's directly to Nano Banana based on that particular thing that you're doing. Where do you where do you separate? I'm curious, where do you separate that right now? What what are some of the things that you're going directly to the edit models for for generation?

Play from 0:44:11 8:88:88 0:44:11 Because obviously it does have a different look, right? A little bit less creative, but when you think about things like text. Or um, I don't know, just other I guess categories. What are the categories that you would lean immediately into on the edit? Very good question. I think It would be things like maybe like CPG, so just a uh like if we're doing a product shot, um, a simple product shot on a scene, simple scene.

Play from 0:44:39 8:88:88 0:44:39 But even with the ones that are like lifestyle images, we tend to go through um Z image. We even try to go through Quen image, which uh that's another model that we use for certain things, certain areas when it comes to lifestyle shots. Um and then it goes to the editing. We we try to actually have the editing model go last because the more you edit like an actual image, the more it crushes it. degrades. Yeah, that that drives me nuts right now. And then you get too far down, you're just like, you gotta go three steps back now.

Play from 0:45:07 8:88:88 0:45:07 Have you God, that sucks. I hate that. Um what about tell me more about Quinn? What are you using Quinn for? is Quen Quinn's also the one correct where Rory, this is what we were talking about last week with the different camera perspectives. it does really well. Is is that primarily what you're using it for. What else are you using Quent for? What have you found it to be great at?

Play from 0:45:34 8:88:88 0:45:34 Yeah, quite Honestly like it's it's certain things like composition and for certain lifestyle shots. Um, if we want we don't use it as often. We used it before Z image turbo came about, um, way more. But it would be things like so there was there was two ways we used it. So the first way was like for Z image Turbo, we'd actually train a Laura on Quen. It had the quality you can actually train uh aesthetics. So uh what we would do is we'd get like images that we like the style of from Midjourney and then put it into Quen.

Play from 0:46:05 8:88:88 0:46:05 do like a styled Laura and then use that and that way it has you don't need to think about the styling or the aesthetic and it did a really good job with that. Um and then obviously go through the editing. But right now we don't use it as often um because we're we're able to achieve quite a lot of that with the image turbo. and then the second way was just for like composition um It is prompt hungry, um, if we're gonna like try and style the scene.

Play from 0:46:32 8:88:88 0:46:32 But the nice thing is you can reliably like I can reliably reuse a prompt and actually get exactly what I'm looking for. Um so that was the other way we used that. I would love to commend you on your use of Laura. That's a very smart area. If you're isolating like a whole process of like where a Laura's gonna work, probably the best, right? Or like the investment is worth it to do one, that feels like the right way.

Play from 0:46:57 8:88:88 0:46:57 Cause I there's a I mean, maybe you have. Have you used a product Laura in a year? Or is it just like totally useless to me at this point? At at this point I think like you d you don't need a product aura. I don't think so. I think yeah. Yeah. oh that's a very like I love that way. Like I haven't thought about the style Laura in that way for like the ideation standpoint, 'cause that's like a really smart like little injection to the whole process.

Play from 0:47:27 8:88:88 0:47:27 So like I 'cause I've been I've been sort of like struggling when people ask me about Laura's and I'm just like I unless it it feels like unless it fits the right part of the process, like it doesn't really ha it's too much effort to for the for sort of like the back end reward on it. Um, and that might just be me, but I I I've been trying to f I still try to find the uses for it. It's kind of like, eh, I don't know.

0:47:54Style LoRAs for brand consistency

Play from 0:47:54 8:88:88 0:47:54 Actually, um so one of the reasons I actually we're using style lauras quite a lot was for brands who have a specific style and they have they have like a very strict specific look that they are trying to accomplish and if they don't want to use mid-journey, right? So that is a that's the that's the only time that I would actually go ahead and use comfy and train a Laura on Quen and you know do the whole pipeline.

Play from 0:48:20 8:88:88 0:48:20 But But I would say right now, you'd be surprised there are like brands that still don't don't want to use Midjourney. Um and that's the only time like we that's what we kinda keep Gwen as a model on our platform. Are the brands that don't want to use Midjourney curious here just mostly larger brands? Or yeah. Yes. Mostly larger and then also um you know, they have issues with the copyright, so they Seems like smaller businesses agencies don't have a problem.

Play from 0:48:49 8:88:88 0:48:49 No, and then yeah, I I with a a a built out legal team is not having it. No. And then the other thing is of course that you can't just only use like Chinese models either. So there's also another like barrier. Um so you kinda have to do the Laura and then, you know, adjust things with nano and then, you know, the are a lot of your workflows built via like like Claude as like the command center via MCP to to foul to sort of execute them or are you still using Comfy a lot?

0:49:11Claude, MCP, FAL, and leaving ComfyUI

Play from 0:49:23 8:88:88 0:49:23 we're moving away from Comfy. We used to use a lot of Comfy. Yeah. We I mean Comfy is it still has its benefits. Um, but right now with Crawd, you know, I'm finding less reasons to use uh Comfy for sure. It just making it like the central hub to do what you need to do. And you can build the workflows in Phoul. I mean it's like the same thing. It's just, you know, probably a little bit more expensive depending on what models you're using and whatever margin they take.

Play from 0:49:50 8:88:88 0:49:50 Um but the it's it's interesting because I think a lot of people haven't necessarily I I'm saying this for for not for not everyone listening here. You guys all know 'cause you listen to this kind of stuff. But like a lot of people haven't figured out like the vector system. of Claude as to how it can be utilized like this. Like it doesn't need to be, you don't need to go find individual API keys to pull in these workflows to create up. You can find a place like Foul where it can do it all in one hit.

Play from 0:50:16 8:88:88 0:50:16 And it can go and search through all this stuff via MCP. MCP like to the to the brain, right? It feels like that's the that's sort of like the the next evolution of this. Cause I'm I'm curious, I saw I saw one of your posts on uh on X about about Fable. And just like it was it was a sarcastic post and I was like, someone's feeling the same way. It's just like they're they're it's so ingrained in the business now. I feel like you probably feel like this too.

Play from 0:50:42 8:88:88 0:50:42 I'm I'd correct me if I'm wrong. But it's like you're so invested, like I would never be invested in like a tech company. Like I don't really care what they do. But now it's like now that it has like direct results on me, I'm like, what are you guys doing? Like what is this? When is this coming? is it Are you finding a bigger difference? Are you finding like uh with fable than like Opus four point eight or like Sonnet, things like that?

0:51:07The model that cut a day to 15 minutes

Play from 0:51:08 8:88:88 0:51:08 'Cause it it's interesting. I mean fable is just another level, right? Like I think Opus you're spending a lot of time And wrestling with the model. Whereas when it comes to fable, it kind of not not reads your mind, but it kind of understands the nuances, understands what you're trying to do, um, and it's more about directing it. Um, and then also like so I like to use it as like a research or like the first bit and then kind of use, I don't know, codex to execute on it.

Play from 0:51:33 8:88:88 0:51:33 But Opus, you know, a task like just for like an example, I did one email template one. I mean, I I think I like literally did that in a couple of prompts. It understood what I wanted to the point where it even did the uh the email where it added links to the social media where when you click on it it actually went to the I was like wow this is fantastic but when I tried to do it on opus I mean it took me a whole I think a day and a half but with Fable it's like 15 minutes it's incredible it's absolutely amazing it is quite expensive but the way I kind of I guess the way right now what I'm experimenting with is like loops.

Play from 0:52:11 8:88:88 0:52:11 ah So kind of having Fable kind of do the research search bit and then kinda like go down the pipeline and give some of the tasks to Codex and then also do the whole looping thing. Um that's what I'm experimenting with right now but it's just it's it's another beast. It did it write did it write the email in HTML or was it like fully designed in like Figma or something like that? It was, I think I was in HTML. It designed the whole thing on brand and everything.

Play from 0:52:36 8:88:88 0:52:36 I think I just kinda gave it like one image of summer Fridays and then kinda gave it the concept and did the whole thing. I couldn't believe it. For for anyone who's not familiar with email design, it's a very interesting because you're defa you're designing like an entire scroll, but it's not it's like basically all one image with components, right? Like this is, you know, a lot of times what you're doing is, you know, you have to you have to design it for the scroll specifically and what you're seeing, and there's a lot of things that go into it naturally like a website, but it's it's complex, right?

Play from 0:53:07 8:88:88 0:53:07 So the fact that it can do this from a reference image is pretty impressive. this one was my first attempt and this was with cool design and that also Yeah. this one, I think yeah, this one also smashed. Yeah. I mean, cause I feel like you did this right after design dropped too, where this was an entirely different use case that people hadn't seen visually yet. And this was like the first time I saw I was like, Yeah, that's dope.

Play from 0:53:34 8:88:88 0:53:34 Yeah. just need the CTA above the fold there. That's it for my email marketing side. That's all I'm gonna say for my email marketing days. But the I still remember. I still judge everything that I see that comes in my inbox. Um but it's like this is a great again, a great piece. And if yeah, there was had it. That one, yeah. That was the one. Yeah. This one was Claude Mythers.

Play from 0:54:00 8:88:88 0:54:00 It was I couldn't believe it, like it it literally just did the whole thing with a few prompts. Uh even the colours. It's interesting because this this one has like the brand colours, whereas the other one had more clawed colours. So it's remarkable. Mm-hmm. And this one was this one was fable and not through claw design? Yeah, this one's fable. Okay. This is so again, it's just like there's there's so many uses for things that this is w this is another piece, right?

Play from 0:54:29 8:88:88 0:54:29 Like again, we're looking at email, go back to my previous statement. We're looking at and I say this only from experience, seeing a billion email marketing templates and whatever. It's like everything still looks like email marketing. I'm waiting for someone to sort of push it a little bit. Like what's it gonna look like? What's it gonna look like when you get real, you know, when uh real dynamic in terms of like the the look and feel of it because everyone's always gone by sort of like the same sort of templates, flow, structure, like nothing ever changes.

Play from 0:54:57 8:88:88 0:54:57 It's sort of, you know, you see a unique one here and there. but it's always like the same visual template, not the same content. Uh not the same visual template, but there's there's like, you know, probably forty or fifty sort of ways to stack an email that'll convert. But it's That was a great use case. I w also, you know, a little applaud on that one. Great. Just finding holes. Cause the other thing that I wanted to ask about, I've hitting I'm hitting systems, I'm hitting all this stuff.

0:55:13Why AI content stopped feeling special

Play from 0:55:27 8:88:88 0:55:27 Content. Like, do you f do you feel differently scrolling your feed now and like what you post? 'Cause it feels like everything like everyone is just spamming AI. Like do you feel like I should put I can wanna post this, but like whatever? It's like it's not it doesn't feel special anymore like it used to. Yeah, I mean that's the thing, right? Like right now, like as you said, a lot of AI, I mean, some of them are good use cases, but then there's just a there's just a sheer volume of AI content.

Play from 0:55:54 8:88:88 0:55:54 And honestly, like it's really hard to come up with like new ideas and stress test things that hasn't already been done um before. Um, but yeah, I mean that is definitely I guess like a disadvantage when it comes to AI. Like it I mean it's good and bad because you know, it opens barriers, anyone can do it, but it's also like I think for creators who normally like I like to kind of be a bit intentional and think about what I want to post. I do find like sometimes when you know, should I just post for the sake of it or should I actually invest time?

Play from 0:56:23 8:88:88 0:56:23 Yeah. How yeah, how are you guys finding it? Do do you want a real answer? I I probably scrapped ninety percent of the stuff that I'm doing because I'm like, i is this compelling or interesting? Does anyone care? Like or am I just posting the post? Like I think there's like a like I I I've gone back to sort of I had like a two month run where I wasn't posting much 'cause I was just so buried and I had no ideas and I was just like, whatever.

Play from 0:56:51 8:88:88 0:56:51 And now I'm back into it thinking I'm like, do I just go with the the volume route and just see what picks up again? Or do I like drill in on something that's a little bit more technical? and like try to, you know, really stir the pot a little bit. But I've been having a tr I've been having a hard time. I've been the first one to say it. I see so much good stuff too. Yeah. to post now too. Like with and it it's probably because we're we're all busy, number one.

Play from 0:57:18 8:88:88 0:57:18 And I think like too, y we we all know this, but it's the best content is usually educational, right? And the truth is the educational content takes longer to create. You know, even with AI, it's it's it's less time than it was, but I mean you still have to be very thoughtful about it, right? I mean you And it could require you recording your screen and then clipping things and then providing, you know, the step by step and doing this and doing that.

Play from 0:57:45 8:88:88 0:57:45 And it's just like, do I have, you know, realistically a couple hours to do that? You know, and is that what I want to do with those two hours? And I think that's ultimately in a lot of ways what it comes down to for me, versus just like, you know, I think the other thing is. I I I wanna also share things that uh don't feel like there's a high effort involved, but still give insight into me or what I'm doing.

Play from 0:58:13 8:88:88 0:58:13 And maybe I know that that's not the best format for that to pop off, but I still want like there's still value in sharing it, just from the context of, oh, Drew's doing this or this is the kind of stuff he's leaning into. And so yeah, it's an interesting trade-off. I mean, Rory and I both too, we used to to do these LinkedIn carousels that I mean, ev every single one of would crush.

Play from 0:58:39 8:88:88 0:58:39 They'd crush every time. So you so you knew, even though you're putting in the time and would do well. But it's just a man, it's a beast. I I I haven't done one of those in six months to a year, you know? I know. I know. yeah. Exactly. Bring it back. so much easier than doing in whatever we were doing. You were doing it in Keynote. I was you know, I was doing it in Canva and Figma most of the time, but now there's MCPs there too.

Play from 0:59:08 8:88:88 0:59:08 It's like Did I mean I d I still don't know how you did it in Keynote. It just it's a very unique process. But the the other thing I've noticed with like the stuff on social now too is it's just like there's like your there's like a there's like a couple buckets, right? And it's just like everyone piles on and that's just like, you know, that's like everyone trolling and having fun.

Play from 0:59:37 8:88:88 0:59:37 Then there's like the then there's like the people that are doing still like the technical educational stuff. And then it feels like the other side of it is just like, let me just create and throw it out there and see if anyone likes it. I don't care. I I I want to be in that bottom bucket, but I like But at the same time, like it doesn't It's hard because I I have been doing that as well and it just doesn't hit. I I mean I've seen your stuff, Rory. Like your stuff is amazing. Like it's all technical, the tutorials, you know, every time you post, I definitely check it out.

Play from 1:00:06 8:88:88 1:00:06 But as you said, like it just takes so long. Um it just takes uh yeah, I mean it it it just content creation takes a while. Um But also like I saw your one, um, Drew, yesterday. Is it yesterday or the day for like the Midjourney? Those were really good. The images that you posted. Yeah. I mean, like, you know, stuff like that, even right? It's like that doesn't necessarily take me a long time. You know, how how valuable is it to other people?

Play from 1:00:32 8:88:88 1:00:32 Probably not that valuable, but I guess maybe it does shine some light into my taste. That, you know, um, what's capable with the tool, stuff like that. So I mean it's limited, but again, it didn't take me a lot of time to do it. And, you know, we're all generating so many goddamn images that no one sees. I mean, that's always been the biggest problem is like I've I wish I could actually just license out my Midjourney archive.

1:00:57The value trapped in AI archives

Play from 1:00:59 8:88:88 1:00:59 Honestly that that is a good idea. Just give me six figures for it and I'll walk away. I'll literally walk away. I mean, I've got so many good images in there that are like even stuff I've generated where I'm like, that's not really my taste, but I know for sure this feels great in another space or for somebody else. And I just, yeah, it's an interesting thing because there's so much value that's trapped in that archive.

Play from 1:01:27 8:88:88 1:01:27 Yeah. And that's a that's another again, going back to just like posting these stills, it's like yeah, it's probably not the sexiest thing, right? It's not the best format, it's not usually not gonna pop off. Um, but I do think it's important still to share. Because we we are creating so many different things creatively, who who's gonna see that if I don't post it?

Play from 1:01:53 8:88:88 1:01:53 You know, like literally it's just gonna collect dust and it's good. So why not? You know? Um yeah. even see it. I'm so curious, do you guys create for yourselves or for the audience? Always myself. I I A hundred percent. You can't it's too hard to try to create for someone else. I don't know what they want. You know, like it's question though. I I I do think people definitely there are people that create just for the audience.

1:02:06Create for yourself or the audience?

Play from 1:02:22 8:88:88 1:02:22 A hundred percent. I am almost into that Rick Rubin belief where it's just like you create whatever the hell you want and the people that like it are gonna be sort of like your your biggest followers or fans. Like there's leverage there and just being true to yourself. The things that you like, you know, and I think that's what that was that's what makes us unique too. I mean Yeah, I I try not to even think about the audience thing. Mm-hmm.

Play from 1:02:48 8:88:88 1:02:48 I will I definitely bread crumb though. That's a hundred percent. Like I I'll do a workflow where it's something ridiculous that has nothing to do with the actual workflow. But like the idea is I want someone to see the workflow and that it fits their need for s something. But like but like make the content absurd. You know what mean? So it's like the the breadcrumbs. Yeah. a Rory Flint thing, for sure. structure. uh Like here's the you know, like the stuff I did with Barkbox where it was like, here's you I'm I'm doing a two D vector to three D design with Cartman as a plush stall.

Play from 1:03:21 8:88:88 1:03:21 It's like it's Cartman, but the idea is like this is for plush toys. You know, so it's like th what make it a little you know, make it social, but like also there's a purpose to this. Um 'cause it it's like how can you make like systems, it's so boring. Like when you when you not not Maybe AI, but like operations, systems like that, unless you're super technical, you're like, I what is this nerd talking about? Um, but like you gotta you gotta like f spice it up a little bit.

Play from 1:03:47 8:88:88 1:03:47 Uh no one wants to hear me talk about system prompts for an hour. It's like the most boring thing of all time. Cool the result's cool, but you know really one of the best at that. I hey, hey, Salma, I'm curious about this. Um, not to put you on the spot. Curious of any hot takes you got. What are you like, is there something you're thinking about that you feel like most people would not agree with or haven't thought about or would push back on that you believe to be true?

Play from 1:04:20 8:88:88 1:04:20 uh Ma maybe not objectively, but like in your mind you align closer to that. Yeah, I mean I think like the whole taste discourse is interesting to me. So I guess my hot take is like I don't think technical people have any taste whatsoever, like software engineers or like any like it's just I mean, I'm sorry, but like they just don't have taste.

1:04:31Can engineers manufacture taste?

Play from 1:04:46 8:88:88 1:04:46 I think the people who come with that taste is people who like, you know, like you guys been experimenting, been crazy for a while, because I do feel like you can build that muscle. But takes time, right? Um and I just see what I see right now is like like I said, like software engineers talking about taste. And I'm like, you're better off getting a stylist who's never used AI. I think they would actually have better taste if they just figured out how to use AI, right?

Play from 1:05:13 8:88:88 1:05:13 So that's my hot take. Like I think if you're tech if you're a company who's trying to push just like, we we're trying to make a model that has taste or whatever it is, um, then probably best to work with the stylist. Who has never used AI, so Mm. you are you referring to that one one company who's who's you know claimed to have manufactured taste? Yes. I that one annoyed the hell out of me too.

Play from 1:05:38 8:88:88 1:05:38 Like it's like I started like scrutinizing things that I shouldn't be scrutinizing. I was looking at like their team photo and I'm like, who took this? Like the c like the it's in the worst composition ever. It's in the corner of the room. Like people are everyone's in one color and one other person's in the wrong. Like, who let this this is like untasteful. Like this is this entire photo. I wanted to like a little mean girl. about? Who are we talking about here?

Play from 1:06:04 8:88:88 1:06:04 Give me the first letter. let's see. I have to see who I have to see who it was. It's is it called taste dot com taste dot com or something like that? um Yeah, em it's interesting because like the stylists and the traditional people are like I don't know I don't know if they get on with the labs but for some odd reason, you know, they don't want to work in AI, but those are the guys who you know these labs want to bring on.

Play from 1:06:31 8:88:88 1:06:31 Like that those are the type people. Is it taste labs? Is that that's it. Okay. Uh-huh. Yes, manufacturing taste. And uh you know I'm looking at I'm looking at all the things from their page and I'm just like, this isn't this doesn't scream taste to me. You know, nothing really screams taste. It's there was I I guess I'm in the I I d I didn't see anything from them.

Play from 1:06:59 8:88:88 1:06:59 So There was one post that went super viral and then it like created a discourse around taste and you know, if it can be automated. oh you know, I I really just when it it's things like this, you know, like when you wanna rebel. It's like the rebellion inside of me comes out. It's like when someone claims that they can do this, I'm like I I don't want you to do it. You know you know what I mean?

Play from 1:07:24 8:88:88 1:07:24 Like I don't even like I don't I don't care if you can, I don't want you to. And if you say y you have and it works then I don't wanna use it. It just like it gives me like an ick feel. Yeah. And then the other thing is like with tastes evolving, you know, like it's not stu it's not something that's gonna just stay the same. So it's just I don't know, it kinda that whole discourse is kinda like a little bit slightly irritating. types of taste. We talked about that last week too, Rory. Remember, we were just talking about like, look, we're, you know, there's stylistic taste, there's concept, you know, conceptual taste, there's cultural taste.

Play from 1:07:58 8:88:88 1:07:58 There, you know, you brought up the great point last week with timing in its association uh association to taste. Like taste is kind of being bucketed and blanketed into this one thing that it's you know, it's it's a deep, variable thing. I mean, I don't think and I think it's Right. It's clickbaity. It's sexy to kind of throw taste into this to generalize it when it's when it's not. I mean, there are just so many different layers to it.

Play from 1:08:26 8:88:88 1:08:26 no, I hadn't heard of this. This is interesting. Now you know, and now you won't now you won't not see it when you scroll by. I Salma, I've I have one more question for you. Um, where do you get your inspiration? and what I mean by that is um Either for new ideas, concepts, um, or things that you're going to try to test.

1:08:38Finding inspiration outside the feed

Play from 1:08:54 8:88:88 1:08:54 I mean, I would imagine if you're like us, X is a big part of that, just because you've got this pool or ecosystem that is so curated in the space, right? So it it feels like a very organic and daily way to feel like you're a part of the social fabric of these conversations. And so there's that. But I mean, like um maybe. Unexpected places or something that maybe people aren't thinking about.

Play from 1:09:21 8:88:88 1:09:21 Like I think Pinterest is a like another obvious one that people probably go to to get inspiration for. But like is there anything else that that you rely on or you've found has been really useful for you? Travel, like traveling to small niche places. I went to Montenegro recently, a town called Coton. There's a little town called Co. my goodness. It's like the there is a mountain and inside the mountain there's like a whole town.

Play from 1:09:50 8:88:88 1:09:50 It's actually remarkable. Um so I like to kind of like find very like unique um like towns to visit. And I find myself like seeing things that I wouldn't otherwise see. So I try to expose myself to different cultures. Um and I try not to look at the obvious places, you know, very niche uh areas. And the other place as well was the other place that I went to as well is like south of France, but not south of France, where like there's these little tiny islands in south of France that no one has like ever heard of.

Play from 1:10:21 8:88:88 1:10:21 And again, like there was a town, I think only like forty people live there, and there's also another one in Vigo, Spain, like a small town and then you know, I like to go to these types of places to just kinda like expose myself to different cultures, see how they think of design, how they think of like, you know, what do their photo shoots look like. And I find myself coming up with the best ideas when I do that. Um and then the other things is like trying to I try to kind of like, you know, go on social media but I feel like I just get the I try to find niche, very niche like accounts on Instagram and on YouTube.

Play from 1:10:56 8:88:88 1:10:56 And then I just save them and I just like bend watch and and it will be like adjacent. It w it won't be things related to necessarily design. It'll be like maybe food or things completely different. And then try to like mix the ideas together. Love that. Great answer. That is awesome. I I'm gonna have to look up this town in Montenegro now. But No. No, I missed that one somehow. I'm getting peppered with that on Instagram and it looks so cool.

Play from 1:11:24 8:88:88 1:11:24 And um Rory, I think it's like you know, it's definitely like on the cheaper end compared to to bigger places for the beauty. And it's probably less of a touristy attraction at this point too. It looks super dope. Yeah. that sounds wonderful. they have these little islands where there's just a church. Like it's it's crazy. Like there's a tiny island and there's just like literally just a church. And then you can the Greek islands. There's like all those little churches on these little islands every That's the so another song, another question.

Play from 1:11:56 8:88:88 1:11:56 This is just a lot this is the last one for me. This has nothing to do with AI or anything. Do you have a favorite nineties or two thousands movie or television show? That is a good one. Um, nineties. I know where you're going with this, Rory. That's right. That's right. There are a bunch like um the one honestly will be like the old I know this might sound so niche and crazy, but like it's the old Cheka Chan movies.

1:11:57Jackie Chan and thumbnail fuel

Play from 1:12:29 8:88:88 1:12:29 Yes. D awesome answer. Yes. was I don't know I I don't know what I was expecting there. I was not expecting that. Okay. Okay. are those are really good. Like the s the fashion styling on those are remarkable. Um just the Yeah. Mm-hmm. that's a great answer. And I'm so I'm so happy that you did that. favorite Jackie Chan movie? Um, there is one.

Play from 1:12:56 8:88:88 1:12:56 It was one of the old ones. It I think it was in the eighties. I can't remember the name though. Um, let me see if I can find out. But it's one of his first debutes. Um Wasn't there some weird there's some weird background too with Jackie Chan, didn't there? I I like read something a long time. I can't remember exactly what it is, but it's like his I think it's connected to his wife or his father in law or maybe his parents. There's some really interesting connection there.

Play from 1:13:23 8:88:88 1:13:23 So it's called the big roll. That's the eighties one. Which one is it? The A big brow. Big b uh I'm looking I'm like looking up his his filmography right now 'cause I'm like uh you forget how many things he was actually in. Junken Master. so random, right? And then and then of course like Yeah. Yes. And then the other one is like Christopher Nolan movies like in a dark night and just the initial ones, yeah.

Play from 1:13:53 8:88:88 1:13:53 Yeah. just just so you know, Drew and I somehow end up talking about something, nineties or two thousands somehow. And then it somehow makes its way into the thumbnail. So this time, you we're I feel like we're gonna have to let let their our guests drive the guess our guest guests drive the conversation in that direction now so that we can always have some some material for our thumbnails, which is probably the thing I spend the most time on each week. Like design wise. Like thinking about it and crafting it and who even knows if anyone likes him.

Play from 1:14:20 8:88:88 1:14:20 I don't know. I just like to do it. So that's the fun stuff. But are you Googling con are we are we getting you know do we are we gonna get served uh or what's it called? Are we gonna get dinged here for doing conspiracy theories on on Jackie Chan? All right. All right, good. yeah, so his identity emerged from espionage, organized crime, smuggling, war, aliases, and a family separated across decades.

Play from 1:14:49 8:88:88 1:14:49 Yeah, I like I I don't want to spend time poking through this. Uh, but there's some interesting stuff like with his with his parents and being separated. And then I think one of his parents was like a spy, like some crazy stuff. Yeah, yeah, yeah. Wait, I do remember. It's to do with Hong Kong 'cause they're from Hong Kong. Yeah. Cantonese, yeah. Mm-hmm. This is cool. Now you sent me down a rabbit hole. Thanks for that. you gotta yeah, you gotta look into this.

Play from 1:15:15 8:88:88 1:15:15 There's i and it's cooler than what I whatever I got just now. I just remember reading about that uh somewhat recently. That was a crazy story. Yeah. well, I think that's uh I have to I have yeah, I think I cause I gotta I gotta run, but that was Salma First off, number one, thank you for being a uh sort of like a trailblazer in a lot of senses here, especially on X and like doing stuff that a lot of people weren't doing.

1:15:29Final lessons from a creative trailblazer

Play from 1:15:44 8:88:88 1:15:44 Um and like thinking about it in a way that was like really technical and smart and professional, like from a processing and everything standpoint early on, not just Now like super early and like telling people about that 'cause that was you know, there wasn't a lot of people back then that were just like, Yeah. you've been a a constant force with with posting. Keep it up. I mean, you always have good value.

Play from 1:16:09 8:88:88 1:16:09 And whether it's viral or not, uh, it's it I know it's got really good, interesting nuggets in there too. And I think, yeah, keep keep doing your thing because everything you're doing is great. And we didn't even get into you know, the brand work and the brands that you've worked with, but you've worked with some incredible brands as well. You worked on a Super Bowl ad. I mean, you've done some really cool stuff.

Play from 1:16:35 8:88:88 1:16:35 So congratulations. I'm glad that those things are paying off for you and keep grinding. Thank you so much. And thank you guys for inviting me to this incredible podcast. And I know y you've been we've been trying to get this organized for a while, so appreciate it. And this has been great. It's legends only. So that's that's the that's the you know the the qualifications to get on here. And you know typically what we do at the end here is if anyone is still listening, if you're listening, you you've sat through this you sat through this whole thing and you haven't either done one thing, you'll light, hit a subscribe button, leave a comment.

Play from 1:17:12 8:88:88 1:17:12 If you haven't done any of that, you gotta go tell somebody. So that's the rules. You gotta tell someone random. So it's gotta be maybe like a parking meter attendant. Something like that. That's a new one, Drew. I think I think I think I got that one. I think that's a new one. Um yeah, New York. The ticket central right there. another one might be your your local World Cup fan. I don't know if anyone has, you know, if there's anyone in the in the area for if you have travelers, visitors, things you know, things from around the US, whatever.

Play from 1:17:44 8:88:88 1:17:44 That would be a good one. Uh do you have any you have any fun ones? We we really like tap I'm feeling I'm tapping. to go into all the niche spaces, yeah. you know, you're uh trying to think about things that uh who's I can't even, man. Did a dental hygienist last week. We're really hitting the bottom. Maybe I got my passport redone. So you'd be you know, maybe you're a local State Department agent. That would be, you know, or or local DMV attendant.

Play from 1:18:12 8:88:88 1:18:12 about to say that. There you go. Now we're original. So