Transcript · August 11, 2024 · 14.7K words
Ep. 6: 5 AI Tools To Mix With Midjourney for Intoxicating Results
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.
0:00:00<Untitled Chapter 1>
Play from 0:00:00 📍 📍 📍 📍 What's up everybody. We are back Midjourney Fast Hours. I'm not even going to do the episode count anymore because it does not matter. We're just showing up every week to do this thing. Drew Brucker, Rory Flynn. What's happening dude? Oh man, you know Episode six like you said maybe we maybe we are counting we lied.
Play from 0:00:29 We are we lied It's but man, this is awesome. I love doing this thing. It's just it's fun to talk about this stuff every day We don't realize how much we actually talk about it outside of here. So Yeah, therapy session for everyone. Let's talk about what's going on. V6 dust is cleared. And I think now we got some more insight into what's going on. But, uh, you know, it's been good to have this new model and just like rip down the rabbit hole again. I was gonna say, man, you're like my therapist without paying you, you know, like insurance is covering it. You know, my kind of thing. Like we're just going in, we're going deep. We're getting all the frustration out.
Play from 0:01:00 We're talking about all the excitement. It's, it's a relationship, dude. It's a, it's a mutual relationship. So I appreciate that. And yeah, man, we got a, we got an awesome episode today. We have a banger of an episode because I think this is going to be a very unique episode so far, which is what this is all about. We're trying to cover Midjourney from all angles here. And this is a very, very important one. We're going to get into. some of the best tools that exist right now to stack with Midjourney into your process, whatever your process is, whatever you're trying to create, whatever you're doing personally, professionally, you're likely using other tools in assistance with Midjourney.
Play from 0:01:38 And if you're not, you're missing out, but we're going to cover the five best ones in our opinion, right now, which I think is super exciting. There's a ton to dig into there. Maybe first Rory, we get into some of these updates with Midjourney v6.1 since the last time we chatted. What we know about v6. 2 and we just kind of run through those real quick. What do you think? Well, it's rock and roll through that. I think there's a couple of things and, you know, not to, not to toot our own horns, but a lot of the stuff that is coming out, we have been discussing here. So maybe Midjourneys listening, that'd be cool.
Play from 0:02:11 Hey guys, what's up? What's up? It wouldn't be surprised. It wouldn't surprise me either the way they operate their shadow organization. So shadow organization. That's right. That's right. I love them. So what was it? It was last week. What was it? No. Two weeks ago we were talking about, you know, Silicon Valley security CIA. Now we're talking about shadow organizations. This is just how we roll. wouldn't be surprised if they're, they're in the, on the back end of something over there. no one knows who works there. No one knows what they do. No one knows any of their personalities. So in this world of, uh, you know, personal brand is everything.
Play from 0:02:46 They haven't seen a single thing. I've seen a couple of pieces on the founder. I swear to God to like, you know, Every person I've tried to look up on LinkedIn has either like been active once in the last like five years. It's like they just Like work in a silo or like underground somewhere, you know profile picture zero, you know Yeah It really is. I mean, I don't even think they have an Instagram account. I mean, that's just, you know, right?
Play from 0:03:15 they're breaking all the marketing rules. You know, like the brand awareness, the building trust, you know, the typical way that brands think about all these things. They're not doing any of those, which I think is just, Insane. We talked about this a ton, but it just continues to blow your mind. Here's a hot tip for all the marketers out there do exactly the opposite of what everyone says you should do that typically works the best. So every time someone's like, do this, do the opposite.
Play from 0:03:45 You know, typically that works, that works 100 percent because what happens is everybody just tries to copy everybody else and then everybody's doing the exact same thing. So again, how do you stand out in, in this crowded space of attention? You have to do other things. You have to zig when they zag, you have to zag when they zig all that stuff. You just look at, you look no further than Joe Rogan. I mean, everyone does. Everyone's podcast got shorter and shorter and shorter and his got longer and longer and longer. And who's the number one buy? You know, has basically more subscribers than anyone else combined in the world.
Play from 0:04:24 all of those, all of the ones that are at the top of the charts are typically long. I think there's always a place for both. Both of those things, the long form content, the short form content, right? Obviously there are pros and cons to each, but I think, yeah, to, to your point, it's breaking the rules of what maybe people would speculate is the answer. Right. And I think you can continue these things and, you know, with longer episodes, You don't have to consume it all in one session, which, which could be a good thing. And I think it allows you to go into the depth, which actually makes things actionable, which is what I always look for in a podcast. Anyway, if I'm going to listen to something, I don't want to just hear ear candy about, Oh, this is cool.
Play from 0:05:02 We did this. Okay, great. But how do I do that? How do, how can I take that to work? How can I think about this in my personal life? Like that part of it is what, what's attractive to me. It's like hearing someone's recounting of reading a book versus actually reading it yourself. like, oh, you get like that second hand sort of, here's like the highlights. This is what happened. Here's the plot. Here's the conclusion. But like, you don't get any of the actual context, which is where you find the learnings. So, you know, for all the people that summarize books on, social platforms, it's cool, but you got to go read them to actually get the totally, yeah, there's apps, right?
Play from 0:05:40 Like so, so audible. Okay. Yeah. Let me listen to the audio book. That's great. Then you have these companies like, Blinkist, you know, where they just, they take all the key takeaways and boil it down to like two to five minutes. And it's like, I, it's not the same. It's not the summary. You think you're extracting all the great stuff, but it's just like, Oh, so at the end of the day, it's just like words that I would read on any social post, right?
Play from 0:06:08 It just doesn't resonate with you because you haven't taken that time just shallow attach. Yeah, no depth. that's just personal opinion, guys. No, no offense to anyone who likes that stuff. It's just, I need, I need more. I'm not smart enough to pick up a couple of lines of cliff notes and run with it. So, We like context. That's what we do here. That's why we talk about this stuff for a goddamn hour. So, you know. Absolutely, dude.
Play from 0:06:33 Alright, let's dive into, let's dive into some of these quick updates with Office Hours and what we heard, uh, last week. So, in painting interface is getting upgrades. We talked about that last week. Hell yeah. Awesome. Don't know any more details beyond that. I don't know what that entails. They didn't share that. Also know that there are new functionalities and upgrades coming with the describe feature. I know that's been talked about a few times on Office Hours in terms of like upgrading that. I don't know if that comes in 6. 2 or not. Could be 7, but that's going to be worked on. Also there are some exploration tools that will be added to the website.
Play from 0:07:12 I would imagine that might have to do with things like style codes or maybe even types of imagery that maybe you can sort by, or maybe just some additional ways to go into specific things that you care about. I don't know, but, I think that's super interesting. And then also lastly, just that sync, that full sync that we're trying to do between discord channels and the web platform. They've done some of that with the rooms, but I think they're trying to find an ideal way. To fully transfer that over.
Play from 0:07:42 Do they also say they were going to work on the zoom in? Was that part of the office hours? I feel like I read that. Maybe I'm, maybe I'm just like dreaming it up, but if I heard, I heard that there was a, I think I'm just reading, like reading the cliff notes that there was some zoom in functionality coming. that would be. Awesome. If we could actually do that. So I know we, we talked about that too. Like, I, I don't remember that from the office hours. I remember bringing that up when we first talked about it on an office hours. And then, you know, David, I think was answered it very quickly as he was going through questions and just said something like, well, why would anybody want to zoom in or crop, you know, and it's just like, well, if you want to pass by that, you know, that quickly, because like to, to, again, to stay in the tool and further refine your image without adding a bunch of additional.
Play from 0:08:28 Post processing sliders, for example, doesn't seem like that big of an ask, you know what I mean? And I think it would be super effective. So that might be the case. I just don't know too much about it. because like, if you have a good image, right. And sometimes there's, there's like one or two things that are really off base and in paintings, not going to work. Cause we know that in painting doesn't work that well when you're trying for certain things. So I might want to crop it out, but also when you crop it out, you have to take it off platform and then. From there, potentially you have to upscale it. And then from there, you know, potentially I have to do some other post processing.
Play from 0:08:59 So to me, it's just like, it's just another feature to like, keep us on the platform. And a lot of the things that we're going to talk about today. Like legit, it's all things that Midjourney doesn't do that potentially, you know, you have to do elsewhere. So these things, you'll see it. They'll probably end up being features down the line Like you just think about that before they added that feature. We had to take it off platform to upscale.
Play from 0:09:26 Same thing within painting. We had to take it off platform to upscale and a lot of the tools that we're using now and comboing really, that's because those features don't exist there, but if they did and they were good, it wouldn't have to go somewhere else and pay for another tool. So not saying. But not dissing Midjourney in any way, shape, or form. They have to develop at a certain pace and not all of it can get done at once. But no, we all have different needs for this stuff and that's why we're going off platform. So I think it's, I think their roadmap pretty much at this point lines up with a lot of these things that we're talking about.
Play from 0:10:05 you know, they keep mentioning secret projects are underway too. They're not going to reveal those. They said, maybe we'll reveal them. Hopefully we'll reveal those by the end of the year. So that's still another, what, four and a half months away until we maybe find out what those things are. That's a big, that's a big question. Yeah. Yeah. I mean, Who knows, dude, I think the end of this year sounds pretty exciting, you know, because you could potentially have these seven, you know, you're talking about additional four months of upgrades that are coming.
Play from 0:10:35 It seems like to your point, and we mentioned this, let's start upgrading some of the tools that already exist. Yeah. I'm starting to see some of that come back that that's been a while since they started focusing on that again. So I'm super excited about that. The, the other quick things let's just run through before we move on is, around six, two, so they're just talking about that in painting and zoom out capabilities, being in sync with six, two, six, two is going to have its own, unique take in terms of like enhanced coherence.
Play from 0:11:07 Text balance default aesthetic. I think they're also going to try to weave some of those personalization aspects into it as well because they're going to have more data. And that leads me to the last key point on that, which is they're going to do another one of those surveys to users. I don't know if you did, did you do that first survey around demographics? Yeah. Yeah, I did it real quick. I wonder if that affected us at the personalization codes at all or anything. I don't know.
Play from 0:11:33 I don't know. They didn't talk about how they were using those just yet. I think the psychometric survey is going to be more based on maybe our upbringing and why we like certain things. And they're trying to, I think, really scrape and assess like why that would be the case and maybe make some, uh, you know, Hypothetical or hypothesis around, Hey, somebody that's this age that came from here that has this background, they tend to like this, Are there any patterns that exist there? So that could be interesting.
Play from 0:12:02 I would be more interested to see how they're going to use that data. Uh, hopefully they don't completely internalize it, but. Or do anything nefarious Midjourney. Right. Like we, like we mentioned, Midjourney is just a psy op, you know, it's just a, it's a big, yeah. Yeah. Yeah. Just to really map our brains. It's how we think and how we want to Kreate, you know, that's, talk about a good sci fi plot right there.
Play from 0:12:33 last two points on this before we wrap is just they're considering a multiprofile personalization system, which I think we all want, The ability to have maybe multiple personalization codes that are used for different kinds of situations. Lastly, Integration of depth control net. We talked about that before, just briefly, there was some thought that that could coincide with six, two. I don't know that to be the case.
Play from 0:12:58 They are talking about six, two by the end of the month. So we're really talking about three weeks here. I don't know how far along there are with depth control net. So that's just a question mark. We'll have to wait and see. Interesting. it feels like a interesting play to have 6. 2, so close to 5. 1. I don't remember what, or sorry, 6. 2. So close to 6. 1 because I don't remember 5. 2 being that close to 5. 1. I feel like 5. 1 came out relatively quick in a comparison to V5.
Play from 0:13:26 And then like we waited for 5. 2. This feels like the opposite. So, you know, I love that they basically train and develop this stuff in parallel though. So like 7's been going on for a long time. So you know, they're just, they're refining, they're doing things, they're training, they're building bigger brains over there. But I guess that's the way you have to do it. You can't just like start from scratch once this new model is, uh, it's totally forever. Um, can imagine what the investment is to train a large model like that. But Hey, It's not, that's not part of this podcast today. Yeah. Yeah. We'll keep it moving.
0:13:31The ChatGPT Aperitif: Priming Your Creative Palate
Play from 0:14:00 Speaking of that, we need to keep it moving because we have to get into these five tools that we're stacking with Midjourney right now that in my opinion are no brainers for some of the things that we're doing. Hopefully it's some of the things that you're doing. You're probably using some of these tools already. So I think the real, the real goal here is to identify. The top five, in our opinion, how we're using them, how you could think about using them. Maybe there are some surprises in there. I'm sure there are. And, uh, you know, really kind of use this as a, as a way to power up your existing capabilities and tool belt within, you know, the AI image gen world.
Play from 0:14:33 So that's, that I think is, is really the goal with that in mind, let's get into number one. All right. Number one, do we have a, no, we don't have drums. Do we want to have like a little drum roll thing? I'm going to add that in here. Drum roll. The number one, no surprise here, ChatGPT, everybody's using an LLM in some way, probably at this point, in terms of who's listening to this podcast, the question is how are you using ChatGPT?
Play from 0:15:03 In combination with Midjourney. So let's start with you, Rory dive in, tell us a little bit about some of the different ways that you use ChatGPT in combination with Midjourney, as well as if you want to focus on a singular one or talk into one more specifically. Totally. I've gone down a lot of different avenues with this. naturally I think everyone goes right to prompt generation. I think that one's pretty simple and usable. Now, again, it's how you do that and how the, Basically to get the best results out of it. Now, again, when you're doing prompt generation to me, you really want to start with a formula because Midjourney will read that and cut and like basically instructions, you can build this to do a custom GPT.
Play from 0:15:41 I remember doing this before we had custom GPTs and I just had like one chat window that was my prompt generator for certain things that I wanted to do, but I had trained it on, you know, basically how to look like a prompt generator. You know, or think like a professional photographer and a Kreative director. And what I want to do is give it a prompt formula. Basically, you know, here's, uh, something, you know, let's call it shot type, subject and action, lighting, texture, composition, camera type, all these different things. And then, you know, give the instructions of we want to use short and punchy, visceral language, you know, we want to think about this, you know, like award winning composition, blah, blah, blah, blah, blah.
Play from 0:16:16 And then you generate and you get, you know, basically you just give it an idea like, Oh, can we generate a, you know, woman walking in Tokyo? Right. And then bang, it would just shoot out the prompt. Run it through. That's a great way to get something done. Now that's very baseline. I think a lot of people are doing that. but you know, there's, there's some more interesting things, right? Like I utilize it for number one for prompt formula generation. Now prompt formulas, we've talked about this before, like these are pretty standard set of like, we need, we know we want to control these little variables and aspects of the image, whether that be, you know, the lighting, the composition, the shot type.
Play from 0:16:54 those different pieces. But to me, what makes Midjourney great is that you have this high level of control. So how can you go deeper into the images that you want to Kreate? So I use this example, you know, if you're doing food photography verse real estate photography, they both have different elements that compose them. So think about food photography, Like what do you actually, what pieces of an image can you actually control? So if you were going to be an art director doing a product shot for, or just like, uh, you know, like an editorial shot. For food What can you actually control on that?
Play from 0:17:29 Everything. So not just the lighting, not just the composition. What's the plating like, what does that look like? What's the garnish? What other types of props are in there? So, you know, look at something like that, where, you know, it's going to be more important to have some sort of. Food photography, again, like how it's cooked, what the garnish is, what the plating is, what the setting is, how that all looks in comparison. So adding those into a prompt formula and structuring it, and then looking up like unique categories essentially for that. So real estate, you might want something like the architecture type, you know, what type of structure or textures are involved? you know, different, different, uh, You know, design styles, things like that.
Play from 0:18:10 So that's radically different than what food photography is. If you think about each category of what you're trying to Kreate. That's a good way to go in there and figure that out. So, you know, been able to build out more robust prompts by breaking down an image into categories. that's one thing to, to consider when you're trying to build these better aesthetics. I think that's a great point where he's not to cut you off, but I think that's, that's worth, really just digesting because there is natural language that exists in various, various languages. Based on those different types of photography. So you just mentioned, food photography versus architecture.
Play from 0:18:47 You know, do you know those terms? And like, for example, how do you instruct using language that Midjourney is going to interpret and understand a certain food to lay on top of a certain food or be positioned in a certain way, right? Like that's different than trying to prompt something that is. Got people in, in the shot, right. Or, or this or that, the other, or an architecture thing. So I think you, there's almost this natural language barrier that exists in each niche, right. Depending on the type of image that you're generating. So I just think that's an important point to point out as you're, as you're walking through that.
Play from 0:19:20 Totally. And I think, you know, you look at it, it's like how everything comes back to breaking images down into words. It's almost like coding, Like you have to code every little piece of an image. This is, guys, this is only if you want a lot of control, and you want a lot of dynamic, sort of Say essentially in what you're Kreating. So, you know, something like that, like you take an image, you break it down, like really into its unique individual pieces, then you can build it back out. And then you have control over every piece. then, you know, even if you don't like something. You can go and easily switch it up. So maybe I don't like, parsley as a garnish and I want dill.
Play from 0:20:00 You can go, you know, label dill as the garnish instead of parsley. Sometimes that's just as simple as, having this broken down into a structure that you can just fill it. So that to me was one of the, one of the big, Better ways that I've used it, to like really find those fine detailed categories that can be super important. cause you know, something like fashion too, right? it might not be what the garnish is. It might be what the jewelry is. It might be what the texture of the fabric is or what type of fabric it is. Cause that makes a big difference. So, That's one thing I'm gonna pass it over to you.
Play from 0:20:35 Cause I know you've utilized this in the same way in terms of like, it's a good segue from the, Basically the regenerate or the, uh, you know, breaking down of an image standpoint, because that's something that, again, why do we have to use that because described maybe isn't the best all the time. Yeah. Right. Yeah. So, I was really just trying to do everything in Midjourney. but then, you know, I think there were certain cases where it just like, Oh, I could actually do this a lot better using chat, GPT as an aid.
Play from 0:21:11 And so for me, some of those things were storytelling, Like, so if you. Are generating some sort of thematic group of images to tell a story of some kind, Which I think is very applicable or practical use of something like Midjourney. Then ChatGPT is awesome for that, right? Because you could say, Hey, chat, GBT, this is my vision for, you know, this group of maybe like, let's just say you created a group of photos first and you said, okay, these three images.
Play from 0:21:42 Can you build a story around these three images and make it blank, blank, blank, and blank. Okay, cool. Now I got a story like this story. I always ask for a few options too. That's always my cheat code. Give me X number of options because you know, that first option may not be great. Plus you can always like. Rip and combine and almost Frankenstein, an ideal option for you too, which I think is, uh, probably an underrated thing, but yeah, you could start with the images or you could say, I have an idea for this, I want you to write me prompts based on this structure, So to your point, if you've got a structure in mind that you really like, take that and say, okay, here's the structure.
Play from 0:22:20 I want you to give me 50 prompt ideas that will. Fall in line with this story arc or this narrative, And so you can build it that way too. And I've done that a few times, which I think is a, is, is an insanely cool use of something like ChatGPT or any LLM for that matter. so like what I can do here is I'll just show my screen just to kind of like provide some visual elements of what I'm talking about.
Play from 0:22:45 And while you're pulling that up, I think the, uh, one, one thing to, to To note here, which was very interesting when testing 5. 2 was, this was how, like, I, I went about figuring out what Midjourney liked at that point was the exact thing you said was giving options, but there was a way that I did it. So I would spit out three different types of prompt structures. I had it structured as one, which was. Super high level keywords that were more generic in nature. So it'd be a shorter, more like punchy prompt. Then I had two, which was basically the way that I like it, which is a lot of details, but like short and concise phrases, and then three was like super long narrative.
Play from 0:23:25 Prompt. So I'd run those three to see which one gave me what I was looking for. And I was like, Oh, okay. I get more control with this one than I do with this one or this one work better. So that was like an interesting way to go about testing this stuff. Um, instead of having to go and, you know, write these narrative prompts that a lot of people do when they, you know, you get these generators to, you know, there's, there's, that's a, that's a great point too, because. You know, prompts, prompts, coherence, and the way that it understands the prompts has, has changed model to model sometimes too, so I think to balance those things out is super smart. This is just a quick example to show you how you could Kreate sort of a cohesive. I guess group of images while also telling a story using ChatGPT.
Play from 0:24:07 Right. So if I just go to this. right here, like all the copy that's in here from ChatGPT, right? I had it right. This series, I then used, uh, this was before S ref codes. So just kind of describe things in the same way. Or maybe I used a, an SF S ref image for this. I can't remember, but obviously they all feel cohesive and a part of that story. They have, you know, sort of that cool bluish green. color to them, And so that's how you can utilize ChatGPT to tell more of the story arc or storyboard and, really, I guess maximize both tools at the same time. The other thing that stands out to me as a way to use it, and I've been using it a lot, is through reverse prompting, is just what I call it, but essentially it's the idea.
Play from 0:24:56 Like describe, you take an image and I created a custom GPT for this. Not hard. Anybody could do this. and essentially what I'm going to do is I'm going to upload this image. so it's going to spit out a very detailed description. Now I've tried a few things, Like to your point, Rory, I could set up this GPT in a way that goes, that gives me three different options, you know, based on prompt structure, like I could do one that is more. like tight, you know, like photography type X, subject Y, action Z, Like all those, as an option. But this one is just a very long descriptive prompt.
Play from 0:25:41 And the way that I sort of set this up is, Hey, you know, can you give me, A description of this image is accurately as possible as if you're describing this image to somebody that cannot see, So it goes into very detailed stuff. Usually it's a very long thing. I was really kind of a little bit unsure how well that would work, you know, in something like like Midjourney, but it actually works quite well. I'll just go ahead and pull up what I got back from that. So here's the, let's go back to the original I did not use an image weight, uh, for this, right? So I didn't then take this picture on top of the prompt. This is a whole new thing and I got this, So you can kind of see how it's a similar theme, but it's a different image,
Play from 0:26:29 right? It's very close to the original, but it's different and you could go in and you could take, you know, it's got the same outfit and things like that. If I wanted to change that out, I could. I went as far as kind of putting Pantone colors in here because there was a period of time where I actually thought they were taking that into account. I don't think that's the case, but it does get you pretty close in terms of like what you're going to get back. and then, you know, like this is super powerful if you want to stack the description with the image, right? Cause then I can pretty much get almost a one for one. Um, let's just take this prompt.
Play from 0:27:05 Let's do this. Let's run turbo. Nice. Oh, it's, oh, it's not going to let me upload it. Okay, of course we're having, of course we're having troubles where we're trying to do a live demo on this, but I think the point is clear. here's the benefit of using it in this way. The reason I use that is if I've got an image and I'm not sure how I want to prompt for it, or I've scoured You know, maybe the explore page or search, and I'm not really seeing anything that already exists that would help me build that prompt.
Play from 0:27:37 This is perfect for that, If I've got an image or, uh, an essence that I'm trying to reKreate and build in real time for myself, I would use that reverse prompting trick. I think that's a great way to not start from scratch to build something. And then just take something that already exists and make tweaks to make it yours. Right? So I use that a ton. I think that's an underrated thing. Have you, have you done anything like that? Oh yeah, that was something that we did for. Basically how we reverse engineered our brand imagery, To, to build out new imagery was we had to come up with a GPT that could actually extract like precise details.
Play from 0:28:13 I can guess, But like a lot of times, like there's some real precision that's needed for, for branding aspects. So, when people were asking me what it was called and things like that, I didn't mean reverse prompting sounds. Perfect for exactly what it is. I just made up the term asset hacking and like, it just sounded cooler because it was a presentation. I was like, I was like, Ooh, asset hacking. Yeah. That's like, so marketer of me, but it was, you know, again, we'd take like brand images and basically, again, if we have our, our structure and what we can actually control within an image, we're talking, like I said, shot type colors, textures, lighting, composition, basically we could take and extract the brand.
Play from 0:28:54 Of the image, we could change the subject and the action, but keep the brand the same. So that was the idea of like reverse engineering this stuff. especially with when ChatGPT came out with the vision capabilities to actually look at this stuff and do it. Yes. Um, yes. And then we, you know, we built a GPT for doing that. It was just, you drop an image in, it spits out the prompt and exactly like what we were talking about before, just like those categories of details, take it and run with it, paste it right in there, go. that was huge. For us when, when we started utilizing it. That way, the other way that's, that's been fairly interesting too, is to like put variations in it from different, you know, different various variables within the prompt.
Play from 0:29:39 I mean, there was a time when we figured out that, Oh, cameras work, like what kind of, what other cameras actually work? You know, you're just like, give me a list of 50 high, you know, really super high quality cameras. And you're like, give me 50 list of 50 Kreative cameras, like what can, what can we get unique visual aspect from? Give me like unique 50 unique photography effects. Give me 50 unique film stocks, Like that stuff all can be super helpful within their like token tokenization. Also randomizing. I think some of the stuff there too is like, you know, I've done this before where my prompt, like randomize the light composition and whatever.
Play from 0:30:13 Like, I want the same subject in action, but I like, don't like this look and I want something else. So like a randomizer effect with that. know, sometimes it's just like, I think everyone probably here has the same problem every once in a while, like we get into these patterns and we generate the same stuff and it's like really hard to go outside of your purview. I personally have that problem. My stuff to me looks a lot of the same because I have a very specific style that I like. And especially if I'm going into a category, I don't know, like if I'm going in, if I want something that looks like it's anime and I know nothing about anime.
Play from 0:30:48 what the types of designs are, like how that's constructed, like what, what theme, like themes there are within anime. you know, I would never know that unless like a, when did a ton of research, but sometimes you can just pull some keywords that might help. So that was such a huge unlock, I think, early on in this Midjourney journey. Which was using it in that way to identify all those things that you just mentioned, like camera angles, composition, expressions, film stocks, right? Like, I think that was like going back to some of the content you and I created a lot last year was around that because a lot of people didn't know to really do that.
Play from 0:31:24 And I think it's, it's even for people like us that are really exposed to some of these terms in a lot of ways too. Like it's much easier to get a full list of those things. Yeah. Then try them out than it is to like, try to recollect those things on the fly and say, okay, let me think of all the different, you know, so, so that's a great way to try and test and experiment with the tool, which is, I think a really great way to up level within something like Midjourney, because you know, what works, you know, what doesn't work, there are certain cool camera angles that don't work, that I wish it worked right.
Play from 0:31:56 So, you know, I think knowing those things. Only makes you a better, Kreator within Midjourney. I think the other thing that stands out too, with like the GPT specifically, I've got a client right now that's trying to build a GPT specific to their brand with my help. So we're like just going through all the camera angles they like to use, all the expressions they like to use all the, you know, demographics of the people they like to use basically, and have that randomizer effect, like you just mentioned, where it's got, you know, can basically just Kreate these randomized.
Play from 0:32:26 Prompts on the fly, taking into account all of those variables across each category to Kreate a unique prompt. That's super cool. there are so many use cases. Lastly, you mentioned this very briefly when we first started talking about it, but sort of like this feedback loop of ChatGPT, because it has that vision approach, you could even. Use it as a consultant or a photographic advisor or Kreative director advisors. Say, Hey, you know, compositionally, how would this look or on a scale of, you know, like grade this from an a to an F under these variables or these categories. What would you rate this image? I actually haven't done that. It's, we should do that, but, but you know what I mean?
Play from 0:33:08 Like you can get to that level. that's why I think this is the number one tool. I don't know that we built this on a, you know, number one means number one kind of thing, but I do think ChatGPT deserves that love because there just are so, so many variable use cases. you can do some really crazy stuff with this. And you know, ChatGPT has been pretty instrumental in a lot of things. sometimes I'll do carousel cover organic social posts, things like that. I'll put the two covers that I'm deciding between. I say, you know, give me your opinion on which one is more visually striking. Break it down and it'll be like, damn, didn't think it didn't even think about that.
Play from 0:33:41 Visual heat mapping. Sometimes it's like, give me your analysis on the heat map of where your eyes are drawn in this. Like that's more for the marketing and branding stuff that we're doing, but you know, it's just like, whenever you have an idea, it's just like, just ask ChatGPT cause it'll help you. I mean, we've also utilized it as a, I've told people to utilize it as a, Kreative partner to like what you're saying where, you know, I typically have like a, I call it, I put this in my course, it's called the three, three P's like, a unique person, unique place, unique position, things like that.
0:34:08Runway's Cinematic Cocktail: Shaking Up Still Images
Play from 0:34:13 so like I want someone that's doing something in a cool place and they're doing, you know, I profession might be the other P where it's like they're doing some sort of like. You know, really crazy job. Like maybe it's a rock climber in Antarctica, something like that. You know what I mean? So like, if I give them that framework and then I say like, interview me, like you're a, you know, like you're a Kreative director interviewing a client and then like start to pull ideas out of me. Like that's a great way to get some stuff like to get some really Kreative images, you know, so you can be A Kreative partner like that. It doesn't have to be your actual prompter It can just be like to generate ideas, you know, so then it's you know, it'll give me like a snake herder in like the malaysian Salt marshes or something like that and i'm like, all right, that's freaking cool Yeah, I would go prompt that
Play from 0:34:59 so not coming to my head initially, but like, you know, if i'm watching nat geo That that would pop into my head as something like, Oh yeah, that was a really cool image. I want to go make that, you know, or actually like take that idea and run with it. So there's, there's no end to it. And I think that's, that's probably the craziest thing about it. think there are a couple of examples that we just gave that people didn't know about. But I also think there are examples that people have used that we didn't say that I would love to know about. So if you guys do have other, I drop them. We would love to try, figure out and just think about things a little bit differently to Rory.
Play from 0:35:38 Before we move on to tool number two, you have anything else with ChatGPT that you want to focus on? Also, only because YouTube says I should do this, if you guys could like and subscribe, that'd be awesome. Shameless plug, doing it right in the middle here. But no, that's the Everyone says to do it. So we're doing it. But yeah, for that one, man, I get there's I could probably go on for a whole podcast on ChatGPT. Maybe we will have to do Maybe we do that. Yeah. Yeah. I like that. So much. There's so much to do with it. but I think, you know, it's just getting Kreative again. It's like, whenever you think you need a tool or something to go look for, like a heat mapping tool, right?
Play from 0:36:16 Like I used to use that for the, for ads where I'd be like, I want everyone's eyes to be drawn to the call to action button. Like, is that if they don't have to go through some super expensive product now, we can just get an analysis, I mean, that's just, it's just so much easier than, or at least comparing and contrasting too. It gives you a much better comparing and contrast than maybe going through a more expensive tool.
Play from 0:36:44 that's a great point that you just brought up to think about the vision part. Of ChatGPT, because I think that maybe is the underutilized part of it right now, because that's still somewhat of a new thing versus how a lot of people were accustomed to using ChatGPT, which is more for the brainstorming, you know, kind of ideation part of it. But the vision component, I think, unlocks a ton of new use cases in itself, some of which we covered. So I think that's a great point.
Play from 0:37:10 Let's get into tool number two, which is Runway. Hell yeah, something my man here knows a ton about because anytime he drops anything on Runway, this shit goes viral, but for good reason, because, you know, Rory's been playing with with Runway for a while now, AI video gen it's getting better and better. It's not quite there yet. And when it does, Things are going to just, man, things are just going to explode. But I think what Rory has been able to accomplish in Runway, even right now has been pretty fantastic. So let's go into that a little bit.
Play from 0:37:50 Tell us about, you know, Runway specifically, how you're using that, maybe the use cases in which you're, you're using that for right now. What's working well right now. What's not working well right now. Totally. Yeah, man. So Runway was a big unlock when it first came out. Gen two was Very good, but it was, it was still like you could tell it was AI, like it wasn't totally polished. The problem they had originally was that you'd upload a super high res image and it would like drop the resolution right away. So you have to then go upscale it somewhere else and they just never looked. Right. So gen two was, was usable. It was usable in certain mediums. We used to make videos on there and turn them into GIFs and that would go into email that would go into social things like that.
Play from 0:38:31 again, it's a gift. It doesn't have to be super precise and super high clarity, especially in small format like mobile. I hadn't even thought about that. Yeah. So that was, that was cause you know, again, I'm thinking about all this stuff. It's like, yeah, video, video, video. It's like, I don't, you don't have to be, you know, it doesn't have to be. Super Bowl ad. Also like B roll for organic social. That's where it worked a lot, you know, very well too. Or you'll see a lot of YouTube channels and things doing it now, like that are storytelling. Maybe it's a podcast or something along those lines.
Play from 0:39:03 They'll have supporting visuals like that stuff works because you don't necessarily have to go look for stock footage. You can make it yourself. so utilizing Midjourney with Runway has been awesome because you can text prompt straight into Runway. And it gives you awesome generations, but it's not as consistent. Like if you want consistent characters, consistent, look, obviously bit journey is top tier from a creativity standpoint. So we can utilize that and then you can also build a consistent look. there's a, there's a few things I'll show you probably a super easy.
Play from 0:39:34 use case for it, because for one thing, like we're always trying to turn assets. Like, how do we take one thing and turn it into other things, how can I take one image and turn it into a thousand, That's how we reverse engineered those assets and then made a lot more from them. The same way we were talking about ChatGPT but not you have these assets. Can these be a static image and a moving image? so how can we take them that turns it into two so a thousand images becomes two thousand assets right now We just have up So like this is something that we would do right here. I'll fullscreen this so that everyone can see it if you're looking at this, right, like this is a still image
Play from 0:40:23 that we created and then we'll animate it like this, and that can be an asset. Like this here, there's a little bit of a, to be completely candid, there's a little bit of a, like, mess up right in the middle. But you can see how it takes this Midjourney image perfectly and then animates it. Ooh, that one's, damn, that one's sick, dude. Right, so this is B roll, this is whatever. This could work as a GIF in an email. You know, we've utilized something like this for ads, like this image specifically, like pulling that into the camera. So you're, that's like what you're seeing on your Facebook or your Instagram ad. So that can be super simple. It doesn't have to be this crazy full on two minute. Cinematic production.
Play from 0:41:01 It can be something as simple as taking, this egg sandwich and giving it some goddamn motion, that's all it has to be. It doesn't have to be something crazy. Now. If you want to go for something that's more crazy, let me make sure you guys might get a pop of a pop of sound here when I, when I expand this, but if we're looking at this, right, like I'm just taking straight up Midjourney images, I'm using my personalization code here, and then we're running it through Runway, Now, This post went fairly viral across a ton of platforms that got picked up. because it's just different than what you're seeing out there. Not a lot of people are trying to do humans. I asked the tool a lot here. When you think about what it has to, what
Play from 0:41:46 it has to actually produce. we're talking about not just human motion, but the cameras moving. We have background motion. We have people moving in the background, multiple subjects. I asked for tattoos. That's really hard for AI to keep consistent majority of the time. A lot of these shots also have a combination of natural and artificial light So when you're looking at that, like right here, we have natural light and we also have artificial light, right?
Play from 0:42:11 Like that could be super confusing to a tool. Also here, same thing. You look at this guy's got tattoos, the, the, you know, he's playing the decks. We got this one. I don't even know how they let me through with this one, but they did. so there's really some dynamic here and you can see some of the motion. This shot coming up here is probably one of my favorites of all time. I don't even know how this is possible. If you look at this one right here with this girl and her hand dancing, just the way that the shadow is even interacting with her face, the fact that that insane, right? This is such stupid. details to notice, but the way that her hand again, her hand in the Midjourney image, right?
Play from 0:42:49 This is interesting hand in the Midjourney image totally fucked up. But because I prompted for motion, It's not like clearly defined now down here You can see the hand is perfect and it's Kreating the shadow on her face and she's got the motion and then at the end Here, you know, we wrap this whole thing up. So Midjourney and Runway to me power couple and What we can do with this and a few tips on this for anyone that's listening. prompting motion into your Midjourney images helps a lot with Runway. If you're utilizing something like a static image, like a fashion pose, maybe to me that never gets the same results in Runway.
Play from 0:43:26 Cause you're asking them, like, if you're saying, okay, here's a, you know, fashion pose and then you're like, you know, put both hands in the air and then like wave them around. Yeah, that doesn't, it's hard for it to do that. But if their hands are already up, And they're already like in motion, like maybe sometimes someone's twisting, like there's the shot of the girl that I was, you know, right in the beginning where she's cheering over the crowd, she's already halfway turned. So it's like, it goes with the motion. So, so you basically need, what you're saying is you almost need to give it some breadcrumbs to work with. The more static it is, the more, Potentially con variables that come into the mix, to mess the image up versus, you know, if you, if you're doing somebody that's walking, right, if they've got that motion of walking in a certain direction,
Play from 0:44:12 that feels like you kind of know what you're going to get along with the text prompting aid. To get the end result that you're you're looking for. Is that kind of what you're saying? Yeah, so this is a i'll actually pull up one more example here This is actually a perfect example of what we're talking about because I think a lot of people try to animate cars Right and there's something with cars that tends to give Yeah, why a problem while you're pulling that up. I was prompting a car the other day. I just got into Runway. he's been telling me, dude, you got to get on Runway.
0:44:17Krea-tive Concoctions: Stirring Up the Perfect Mix
Play from 0:44:44 Got to get on Runway. I've just been kind of waiting for the tech to get to a certain place. I think it is probably there. Now I jumped in to your point, did a car. It was, you know, in the prompt in Midjourney or in the photo in Midjourney, it was just a static image of a car in an alley. no motion blur, like you see here and what you're showing and, you know, tried to put something like that into Runway.
Play from 0:45:16 but the car just kind of slid to the left. It didn't go forward. Like I was expecting it to go. So I say all that to say. it sounds like putting something in motion already in a static image, you're going to inKrease your likelihood of getting the result that you want. A hundred percent. So like with cars specifically, it's going to be a problem. Cause if you have a static image of a car with a very static rim, like it doesn't know to necessarily animate that. So when you're prompting, especially things in Midjourney, hot little tip. You can prompt spinning wheels. So it looks like the wheels are spinning.
Play from 0:45:52 Also, sometimes that doesn't help because if you have, a background that's fairly static and clear, also the car tends to move backwards a lot for whatever reason. but if you can get this motion blur effect, which is pretty much all you have to prompt motion blur. So like, you can see like the, the lines are going. to the right Like they are flowing to the right, which then gives Runway this like little headstart on what exactly you want to do so that when it's going, like it just knows to go that way and the wheels are spinning. it comes down to the reference image.
Play from 0:46:28 And also sometimes when you're text prompting, you know, driving to the left, or, if we go into this example here, these are a little bit more, there we go, sorry for the, for the music, but these are a little bit more dynamic, like you'd have to put in the text prompt, like driving towards us or driving towards the frame, like you want to tell it that because if not, it's just might just start rolling backwards, This one, I think I just got lucky on this next one because it has a track and a path with the road. Basically, I was saying driving towards us on a curved, muddy road or driving away from us. Sorry on a curve, muddy road. And this is how you know, it can end up really driving in hitting that track. Now it's not perfect, but it's good because you'll see there's
Play from 0:47:14 a lot of a instances here where you just end up wasting a lot of gen time because it's hard. So yeah, I think that's a great man. That's a great tip. I wish you would have dropped that tip before I. Exhausted all of my Runway hours in literally 20 minutes. I got to do some more Runway stuff. Like we're actually break it down. It's just so, there's so much content that needs to get made. Sorry guys. I know it's just, it's just a whole nother, it's a whole nother beast. So I think that's, I think that's a great point. You have anything else you want to wrap up with Runway before we move on to the. Tool number three.
Play from 0:47:48 No, just like when you're text prompting and or image prompting, just remember you're working in four dimensional space or three dimensional space, whatever it might be, it's not image like, okay, here's, you know, you're flat. the camera motion is really what is important in that. So the camera motion can go forward. It can go backwards, up, down, diagonal, how, you know, whatever you want. You have to be able to prompt those. So another good use for ChatGPT, if you don't know them, look up the camera motions, shot types, things like that for cinema.
Play from 0:48:19 Versus photography, so you can understand what a 360 degree orbiting shot is, you know, you're really just directing a scene at this point, but a lot of times it comes down to the reference image. So that's, I'll wrap that one up, there. Great point. All right, well, let's get into tool number three. Which is awesome tool, something I've been using quite a bit. I use it for a couple of different things right now. So primarily for me, which this is a little hint into tool number four, but it's one of the reasons it's one of the tools I use for upscaling specifically.
Play from 0:49:00 So outside of Midjourney, I will take things into Krea to get that additional upscale. The upscale part of it is free. I am not on a paid plan. So everything that I'm utilizing it for is just on the free plan. I have paid for it before you do get it. Some additional customization options, of course. But, upscaling totally free. I can show you. So let's Let's go ahead and share screen. We're making this a very visual, very episode here.
Play from 0:49:32 So let's just pull up Krea. I'll go into enhance. Let's just take an image that I've already done. maybe let's just go in here and you basically have these options here that exist with the prompt, the strength of the AI that you're going to use, how much you want it to resemble the existing image that you have, and then the clarity component, those are typically, I think the main, main things you have here, there are some other things like scene transfer, if you want to remove the background, for example, and change that out, but this is pretty powerful.
Play from 0:50:09 I like to go in here and play around to get that additional detail information. So you can obviously see in this example. Image with the clouds. I love the cloud, the shape, but it does lack maybe a little bit of texture, right? Like, and look, I'm not a cloud expert. Let's let me be clear. I'm not a cloud expert, but like, I think from an image standpoint, I want to add some personal, you know, personality into the clouds. And so this is kind of what you can get with these AI enhancement upscalers, add in some of that strength, how much I want it to resemble it. Some of that clarity and then obviously creating a larger file size. So if I'm trying to print something out, this is for a client that wants something like that, I will put it in here and just sort of find wherever that sweep spot is, like this one, the clouds are a little bit
Play from 0:50:57 less. This one I think is sort of a little bit more of the sweet spot I was looking for, where it's got more of the texture, but it's not overdone, right? So this is a nice way to refine that. I use Krea primarily for that. I also use Krea for. I would say like photo restoration. for example, if I've got an image that is very low res, like this, let me just zoom into this, This is a mush, This is a picture. This is a picture of a painting, So this is about as low quality as you get at. And it was taken on like an older phone. So you could do this, Where you just the strength, resemblance, clarity. Turn it into like something like this, So it's just a little bit cleaner.
0:51:37Magnific's Magic Potion: Supersizing Your Creations
Play from 0:51:46 Okay. Now we're kind of getting somewhere and that wasn't quite what I was looking for. So I ran it again. And now you've got a lot more detail here, You've got more detail into the actual strokes of the painting, which I think is an interesting approach You could say, Hey, look, those waves are a little bit sharp. The background's a little bit sharp, but you know, I think that's where You know, maybe even taking this a step further and stacking another tool on top of it, comes in because I could soften the sky. I could soften, you know, the the waves or whatever it is And I've done that before too. But the point is, is you can add those really fine details into an image here. You could see this a and b right before. And depending on how much you want to change the
Play from 0:52:34 image, that's where these sliders come in. So I think that's, that's use case number two for me. And then lastly, I'll just go into generate, which I think is a very, very cool use case. It's basically like, I think this takes shape more in like a product capacity, but this example I think should do the trick. I could take an image from Midjourney. Put it into this generate feature of Krea and adjust this AI strength to my liking. And as you can see on the image on the right, this is just adding details, unsolicited advice in terms of the details that it's adding, right?
Play from 0:53:07 So some of this I could really like some of it. Maybe I don't like, but you could just start to see how you could take this imagery and rather than Trying to describe all those things into an image or just looking for a creative partner in all this. I could just drop this in here and just go from, I mean, it's so cool. And then, you know, the icing on the cake is, let me take this paint brush here and I'm just going to paint something on the shoe like this. Okay. Now it's going to take that into account with the image. How sick is that? Match the texture solid, you know, it's just insane capabilities. And if you're a product designer, somebody that could utilize something in this capacity, or you're just looking to expand on something you've already created, this is a phenomenal tool to do all that.
Play from 0:53:53 And again, This is all just on the free version. obviously you can see this is a extremely powerful tool to have in the tool belt. Rory I actually haven't tried the video, capabilities, It's interesting. It's more like, at least from my experience, I did it for like, I don't know, a couple of days. More, more surrealist in nature. it's not super precise, like anything you're seeing on Runway. It's got a lot more creativity baked into it. So, very interesting in that sense, but you know, I think if it has its own specific use cases, right.
Play from 0:54:36 It has a video upscale or two, which is nice. That's something that a lot of the other tools. I didn't even know that. Yeah. So that's, that's a good little like tidbit there. Cause I think that's what we're me personally. That's what I'm missing. And you know, topaz is something that I use for upscaling videos. It's okay. to me, I never, I have to add a lot of frames to make it actually super, uh, you know, clear, but that also 45 minutes to render. I think it's probably my hardware that sucks. But that's just, I don't have time. I don't have an hour to wait for something to render to see if it's okay or not.
Play from 0:55:13 Cause then if I have to run it again, it's another hour. So I'm like, all right, if someone could, you know, if someone could make it as easy as Krea or our next tool, I would really appreciate it. So if there's any developers out there who want to make a billion dollars, a video upscaler that has a nice, super easy, super easy to comprehend, UI, that would be. Very helpful because you know, you get into Topaz and it's like, I don't even know what half this stuff means. It's just really, I'm not like, I was wondering about that because I did see that Topaz had something out with video upscaling and I just, I didn't know anybody that tried it. So that's interesting.
Play from 0:55:49 in the videos that I saw, I wasn't like overly impressed either. usually you can make, you know, cherry pick something in a, product or a feature that, that makes it look, extremely good. I thought it looked okay. before some of the updates, they used to just have all these different models that you could use. And it was just like, there was no, there was no description. It was like Kronos Kronos fast. I'm like, what does that mean? Like, why, what model am I choosing? For what reason? product hell, you know, probably could have gone and done my research. But at the moment I was just clicking buttons. I was like, let's just see what happens here.
Play from 0:56:31 people ask me about it a lot. So I really wish I had a better one. Krea's does fairly decent, but also feels like it, when you upscale it, Too much or don't have the right settings. It like de saturates everything into, or not de saturates, like softens everything into looking like, you know, 3d renders and not like realistic. I haven't found one. If anyone has it, drop it in the comments, please. Yeah, sure. People are doing it with, uh, You know, comfy UI in some way, shape, or form that they're hiding from all of us. I also dropped those workflows if you got them.
Play from 0:57:01 Well, let's get, let's get into tool number four, which is the yin and yang of, of Krea, right? Which is Magnific. You are using Magnific a ton. A lot of people listening to this podcast have probably used Magnific as well. Does some upscaling probably the best upscaler would you say right now that exists in terms of AI upscaling? I think so. That's just my own personal opinion. Look, I've tested a bunch. You It's just to me, my personal preference because of the level of control that you have and like the different features that it has. It's expensive. Don't get me wrong. I wish I, you know, personally, I love the tool. I wish it wasn't as expensive, but why am I going there?
0:57:25Adobe's Top-Shelf Finish: The Perfect Nightcap
Play from 0:57:39 Because Midjourneys upscalers to me don't do what I'm looking for them to do. And that's not to say that there's a bad, it's probably not a big concern of theirs to worry about image generation, not upscaling. But I really like to get those finer details because I feel like that makes a big difference. With, with something that you're doing in Runway, do you find that you need to take those images into Magnific before you drop them into Runway? Are you doing that? Makes a big difference. Makes a big difference. So the lower the res of the image, the harder it is for Runway to generate, the better it is.
Play from 0:58:16 Like you used to lose resolution in Runway. Gen Gen two. Now we're at Gen three. Gen two. I'd put in a really high res image and it would like shrink it down and like make it all pixelated. But now it feels like it transfers one to one. So I'll show you even like the, that now might not be able to notice all the fine details. But if you look at certain pieces, like the depth of the crispiness, then also like the sauce, it just gets way more textured, Like there to more textured, to get a little bit more light gloss.
Play from 0:58:52 Like these are simple things like the bun up here to zoom into that a little bit more, if you can't see. So yeah, this is great. I think that's where, where Magnific really shines above the rest is just, you could keep zooming and zooming and zooming and it's just insane how it puts it together. Like that is just crazy from crazy. Just like the char on the, on the bun, it becomes so much more defined. And then, you know, let's see here, Yeah, you're looking at the crispiness of the chicken here that goes from like somewhat fuzzy to like really that's again and that to me makes a difference in Runway that that starts to appeal to a different different senses at that point too.
Play from 0:59:35 So when you're talking about food, it's just an image. I mean, food resolution is so important, Because then you can activate other senses like tastes like smell that you can pick up through how clear. Yeah. That is, and I think that makes a ton of difference on, this is a perfect example for that. So this one, cause it's not as close up, Like it's fuzzier. Like this top patty, this is one that we'll focus on. That one gets a total upgrade and looks way more, you know, like food photography. We even get down this bottom one, right? Like it gets more defined. This gets, you know, the lighting gets enhanced, you know, especially on the ketchup.
Play from 1:00:16 Like you just see these little, like. Contrast definitions up here to like actually separate the ketchup from the chicken. And if we even zoomed in on that again, guys, these are really small details, but they make a difference. just the difference of that is big. So that to me is some of the best. Now this is where I think it really shines is putting like a realistic touch on humans, which I think is the best that is out there. Cause if we go and look at the original gen., that's, This person's face. Now it has wrinkles baked into it, but now we really add some natural skin texture to it.
Play from 1:00:50 There's probably some things that I would edit in this photo But now you start looking at the difference here And you see like that, just even like these, these cheek dimples and wrinkles here, the, the layers and the texture of the skin, this becomes so much more. Insane. Because there's a bunch of examples. This one, to me, I use a different level of fractality on it, which means it's going to be a little bit more creative. let's zoom in on her face, Like, you can see the difference in her eyebrows and her forehead, right?
Play from 1:01:31 Like, the difference there between that and that. Like, we're looking way different. Some of the skin, right? Like, maybe I'd make some different adjustments here. On the skin, maybe post or whatever. I look at just the eyebrows itself and just, yeah. You, you, I mean, you could start to actually know how old that person is versus have a, have a general idea without it. Right. I mean, that's, that's what, that's what we're talking about here is those, those forehead wrinkles, right? The crow's feet, you know, like all those components of a face that really are You know, age. So even when we look at this person in the background, they're fuzzy. They get a little bit more definition.
Play from 1:02:14 Yeah. Like that becomes a little bit more defined. You can see the contrast, you know, you can see the shadowing, the lighting. She might have a little bit of a Ripley chin cause we're so close, but this is what I like it for is making stuff natural because not everyone is perfect. Not everyone looks like a model here. Like to me, these are great. Like this one. Define the contours of her face. Sometimes to look out for is it likes to add freckles for some reason, like every AI tool for whatever reason, but you get what you don't get with a lot of the other upscalers is like this, this like blemish here, right? Like this is a blemish.
Play from 1:02:49 Most people would want to airbrush that out. Me personally. I like it in because it's just real. These skin, skin lines are real, gave her a little hair on the upper lip. Sorry about that AI lady, but you know, even like the hair gets more defined over here and things like that. have the capability to make those little healing or spot removal adjustments in the tool? No, I wish it did. That's one thing that I wish it had is like, Oh, let me just do this instead.
Play from 1:03:19 Maybe that's the perfect segue into tool number five. Rory, do you want to go ahead and reveal tool number five? It's it's Adobe. Everyone been trying to get off it for my entire career, but it still has uses. And I, you know, Figma can try to take me away. Everything else can try to take me away. Canva, whatever. I still end up in Adobe. So generative fill to me is just super useful, especially for fine tuning, spot cleaning. Doing, you know, everything you need to do from a, fine tuning. Art perspective is just, it's just, it's unmatched really is.
Play from 1:03:56 Gen. fill to me is, is just obviously a game changer because now you've got two options for in painting. So to speak, you could try it in Midjourney. Doesn't work. Try it in Photoshop. I use it for time to time too, is just to expand the aspect ratio. So like my, my image is a four or five, but I want to create more negative space above the person's head or below their feet or to the side. Perfect for that, I mean, so I, I think that's, that's another one. One of the other things I probably would have used
Play from 1:04:43 it for a lot more before we've gotten to this quality with Midjourney is probably like color grading, adding LUTs, you know, things like that to get the tones I want. But now Midjourney is so good with stuff like that. That's really not that big of a need. The last thing I'll, I'll just say before I turn it back over to you is probably just text. You know, if I, if I want to put text, Somewhere, and I don't want to do it in Keynote or Figma or anything else that I'm using.
Play from 1:05:09 That's an option, right? If I want to put the text behind a person. Obviously like Photoshop's perfect for stuff like that. The composite stuff obviously still got a huge need. masking, right? Like if I do want to make some of those micro adjustments to, dodge, burn, whatever it may be. So some of those processes, yeah. He'll that obviously heals, heals a great one, Like just cleaning up some things. So what about you?
Play from 1:05:36 Yeah, same thing. I mean, look, there's sometimes I like to. To screw around in there and kind of separate layers, So I'll go in and take a, like deconstruct a Midjourney image, you know, like mask, certain layers can basically copy paste things and then add different depth value or different depth into it, or change up with generative fill, like certain pieces of a layer. but you know, the thing that's, that's great about it is you can go down to the pixel on the generative fill. So. You know, that's the one thing with Midjourney, another request is that I can't wish I could zoom in on my photos a little bit to see what I'm actually selecting.
Play from 1:06:08 Cause sometimes the finer details are harder when it's just like a full image that you can't zoom in on, but Photoshop can go in there and change one pixel. If you want to zoom all the way in, which is super helpful, but I love the exactly what it said, like the generative expand background removal, which thank God, you know, they have the, they have the selection tool also, which is like, you know, the auto select, which whoever's in here and, you know, it's still loves manually going and outlining an entire.
Play from 1:06:35 Piece of an image in Photoshop, more power to you. I used to, that used to be the bane of my existence. I hated it. So those simple features, like they're little ones, but they help. And I do a lot of post processing, a lot of times I have to soften some of the images up from, from either Midjourney or magnific like magnific specifically when sometimes I just love it. I know the generation is not going to get any better. It's toss into Photoshop. A little soften here, a little, you know, edit there, so it's like a, it's a whole creative process, I mean, look, you can use all these tools in one workflow and you can use all of these for specific things.
Play from 1:07:14 I did this for the, you know, for the AI for the AI threats presentation that I was saying it was basically, Runway or sorry, Midjourney into, Magnific into Photoshop to soften it up and make it look more realistic. And then, you know, take it into one of the animators and animate myself as a deep fake. So it's, you know, the, the other tools that are out there can do a lot of this stuff. These are just the ones that I think I've used probably the most. and if I had to say, if I'm going to rely on it for. Client work or putting out something professionally. Those are the ones I'm using.
Play from 1:07:49 Yeah. Photoshop almost seems like a must have in that workflow. If you're going to use it professionally, for whatever purpose to clean it up, to post process it, whatever that means to you,
Play from 1:08:53 plugins too, that just came out. I think, krea and comfy both have plugins for Photoshop. The other thing I will mention here too, any designers, this is just a hot tip. If you're on illustrator right now, just remember this. there's a mock up generator feature now, which is insane because that is just up such a pain.
Play from 1:09:22 you can use Midjourney to generate, let's just say like a blank. Model, like a model with a blank T-shirt on, or a blank hat, whatever. Then if you have a SVG or vector file of a logo or a design that you wanna put on it, you go into Illustrator and use the mockup generator and it'll move the image around dynamically on the model. So if my T-shirt's wrinkled. It'll integrate it with the wrinkles. So it doesn't have to be like this crazy process of what it used to be. you can just move it all around. I can move it onto the hat and then it like moves onto the side. So it'll change the dimensions and the warping.
Play from 1:09:59 that is insane. Whoever. Has done this before knows how much time that is and how often it never comes out the way you want it to, unless you're really good, which I wasn't. So that one, keep that one in the back pocket. I like that. that's a good one. I also like that text, That text part of it where it's like you take text in illustrator that you move it over to Firefly and then you can basically just make a simple prompt in terms of the texture or whatever you want that, that text to turn into inside the letters, for example, into a banana or into a concrete or whatever it is, like brands could be using that right now, probably not that hard to do.
Play from 1:10:39 I haven't done it personally, but I've watched the steps. That looks awesome. I need to play with that. That's a good reminder to play with that as well. There's so much you can do with the tools. I mean, it's really just like how creative you can get with all every single one of them. Like I said, we're using a lot of these because Midjourney doesn't offer them in platform. If they did probably just use them there, but it's also good cause you can get to learn stuff. I forgot more of Adobe than I remember. unfortunately, which is sad for me because I used to be in it all the time.
Play from 1:11:11 But, you know, having it this in there now, I just, it feels like I'm riding a bike on some tools where I, you know, come back up. Pick up some of these things and go, but I have to remember how it was trained. But Firefly, I feel like I neglected and I probably should use it more because there's so many men. There's so many, we just live it, you know, out of the, this isn't a 537 AI tool for marketing posts.
Play from 1:11:37 Sorry guys. This is just the ones that we think are useful and we can actually utilize today. maybe my main takeaway is this, Anybody can technically. really good at Midjourney. If you put in enough time, I think the real Asset to have both maybe for yourself or for a company that you work for is the ability to stack things together into a workflow, build some sort of operation.
Play from 1:12:07 actually with all the Midjourney features, the more Midjourney features, you know, and you know, well, and you know how to troubleshoot and reverse engineer probably the better you are at Midjourney in terms of like doing this for professional work, you know, to bringing that into the business fold. The same thing is true with. Workflows and adding these tech stacks together, knowing the different use cases can unlock additional and more powerful ways to use these things, which I think is really, again, I think that's the takeaway for me here is knowing how all these things work, the connective tissue within them, and then how to squeeze as much value as possible out of them.
Play from 1:12:46 Yeah, a hundred percent. also, it's just problem solving, right? Like when you can't do something in Midjourney, there's a tool for it. Like, Oh, I can't upscale this big enough. What, what's the next logical step. Go find an upscaler, Oh, I can't animate this and run in Midjourney. What's the next logical step. Go find one, right? It's just simple things that have to be, you know, also it's like a desire to actually get to a finished product and not just give up, which sounds so stupid. You know, personal responsibility, but it's like, really, like, if you want to get to an end result, you can get to it.
Play from 1:13:20 There's another, there's a tool for that. I mean, you know, you look up, Oh, I can't, I can't turn this logo into a vector file. I guess I should just give up, you know, there's an AI tool that turns logos into vector files, it's there. That they're all put on your, put on your QA hat and literally find problems because there's so many, Yeah, there's so many. And to your point, like the more problems you can present yourself with and think through and work through again, the more you're going to level up and know how to these, how these things can be utilized and leveraged the most effectively. So I created my logo with Midjourney.
Play from 1:13:58 Illustrator, I prompted in Midjourney. It took my logo out. I put it into Photoshop. I cleaned up the background. I took that into, illustrator, made it a vector file, changed up the colors built around it. And then there's a logo, I knew what I wanted. I just wasn't going to go and illustrate it. so, you know, it's just, what do you, again, what are you guys experts at? What are you experts
Play from 1:14:50 at? What do you know how to use? What do you know how to use? Well, What are you, you know, how can you apply this stuff to your current processes and just make things a little bit easier? And your day to day, that's my two cents. Sorry for getting on my high horse there at the end. I love the high horse. And I think as we mentioned, this is a great chance to think about your own tools, workflow. If you've got tools that we didn't mention that you think deserve to be on this list, if you've got workflows that we did not mention, let us know, drop us in the comments, drop this on YouTube.
Play from 1:15:25 Ping us on LinkedIn, whatever it may be, you can find us. Let us know. We'll give you a shout out. We want to think about this thing as innovators ourselves. We're constantly looking for new use cases. So love that. Rory, man, this was another dude. We, we filled this episode to your point, I think we need to do another one on ChatGPT, but, this was fun, man. Anything you want to wrap up with before we, we dip.
Play from 1:15:50 No, man, that was, like I said, I think we should definitely go in deeper on some of this stuff. Maybe we can put together some more workflows and things of that. ChatGPT is endless. the other, the last piece I will, I will console with that is that, um, it doesn't have to be ChatGPT I mean, ChatGPT has the vision capabilities and it's really good, but Claude works really well for prompt writing and for different, you know, terminology. I tend to find different things.
Play from 1:16:16 On Claude that I do on ChatGPT. So if I'm looking for cameras or film stocks or. Any sort of like research, I find different things on Claude than I do on ChatGPT. So that's just something else to keep in mind too. One of the best things I did early on with LLMs was run the same thing two or three times in different tools, just to see what, what results I got back. And I think that's an extremely great point. That still makes a ton of sense to do today. Just take the same thing, see what it gives you.
Play from 1:16:45 Also like for this episode, we're talking about Midjourney, but Claude is just exceptional. writing anything, writing, I'm going to Claude. So, I mean, Claude deserves its love. It's an awesome tool as well. I'm glad you brought that up. That's a great honorable mention. Yep. Honorable mention, Claude. We love you though. Anthropic shout out. But, uh, yeah, guys, thanks for joining us on episode six. If you haven't already followed the show, whether that's Spotify, Apple. Drop us a sub on YouTube. We're growing there. I think we're almost up to 2000 subs now in 30 days, which is just crazy. You guys are just, you guys are animals.
Play from 1:17:20 So killers. Awesome. Love it. We appreciate the support and we'll catch you guys in the next one. Hell yeah. See you later, man.
