Transcription
Most people are still editing their thumbnails and graphics the old way, pixel by pixel, tool by tool. But what if there was an easier way? Today, I've got David Altizer, who was a full-time YouTuber and became a full-time thumbnail designer who works with some of the top YouTube channels in the world. And he's going to show us exactly how he creates thumbnails and graphics that actually stand out using AI tools. Hello, David. Thank you for being here.
Thanks for having me. I'm happy to be here all the way from Nashville, Tennessee. So, thanks for inviting me. I'm stoked to teach you guys how uh how I make thumbnails. And I wouldn't be here if it wasn't for AI. I started my entire career right when Generative Phil and some of these AI tools came out like what is that like three years ago now? Uh at the time of this recording. So, yeah, I'm excited to share my journey and and share my process with you guys. And yeah, let's go.
We are honestly so excited to have you here because anyone who wants to be a content creator has to understand that the cover of the book or the thumbnail of the YouTube video is the main standing point and back in the day it was books. You'd see the cover of the book and you decide if you like it. Now the it's actually your YouTube thumbnails that are the most important thing that are going to get people clicking. So, we are so excited to see your process, to see what's working in thumbnails, what's not working, what you recommend, how you're doing it. I just I'm excited. I can't wait to dive in.
Oh, yeah. It's It's radically changed uh everything. I would say 95% of the thumbnails that you see on YouTube at the highest end are I say 95% AI. So that last 5% is usually a face or sometimes a specific object that just feels a little off with AI. But even then, um, a lot of times, you know, even some of the biggest creators, it's a 100% AI in the thumbnail. It's kind of one of the best kept secrets in the industry, I guess you could say. A lot of people don't talk about it, but um, it's gotten so good now that, you know, it's it's almost a waste of the creators time. Uh, when we're dealing with people, you know, who are essentially celebrities like Mr. Beast or you know some of these big creators, it's a waste of their time to take pictures for their thumbnails. And now we can use tools to basically recreate them so that we can put them in whatever pose we want for our thumbnails. And it's just really been an amazing tool. And uh, yeah, you got to figure out how to make it not detectable though. That's the secret sauce. And it does require human uh involvement to do that. And yeah, I think taste is the main thing. You have to have a a good taste uh and understanding of what looks good, what gets the click, the psychology of the click. And yeah, I'd be happy to share some of the tools that I use to do that with you guys.
Oh my goodness, I'm so excited. Like just even the fact that you said that there's like a kind of little secret source in there as well. I'm like, yes, we're getting all the tea.
I'm super pumped. Do you think do you think you could start by showing us some thumbnails that are good compared to some that are not so good?
Sure. Um, so I've got my screen share pulled up here. Um, this is a creator named Annie Austin that I started working with about six months ago. And his style is he's on a green screen. I'll just click into one of these thumbnails here or one of these videos here. And his style is basically like the thumbnail. And that's one of the key ingredients honestly of making a good thumbnail is being honest with your designs and you know, delivering on what the the potential viewer is seeing. So I'm always looking at the way that the video was shot, the way that it was filmed and kind of matching that energy. That way when when you look at the thumbnail and you click into it, you're not being misled by what's actually in the video. So, you can see here that that Austin records himself. He always has that same t-shirt. He always has those glasses. He always holds the microphone and he's always on a green screen. And so, we take that same kind of overall aesthetic and apply that to our thumbnail designs. And the all of these designs here are mostly AI but mixed with real assets. And uh for the most part, all of those images of Austin are usually photos that he takes on his iPhone. Um, and that's one of the really big secret ingredients um I feel with thumbnail design.
Here's another creator I work with, Noel. And we do the same thing with him where we have all of the elements basically like you can see here are AI generated behind him but then the actual face the reaction face is a real photo and uh that's not always the case like in this one here the teaching a one-year-old how to use prediction mark markets um that one's a full AI body so it just depends but I found using little human elements ments in the thumbnail helps add a lot of that human touch so that your viewers and we've actually had situations where the viewers of these channels call out like why you ew ew gross this creator is using AI gross you know cringe right AI slop so to avoid the AI slop factor we were always for these creators at least using um real photos of their faces so we could I can show you my process of that in a minute. But those are just a couple of examples of that. Um, there's another creator that I work with named Curtis Connor. I don't work with him anymore. I actually taught him how I spent about two hours with him on a Zoom call and I taught him how to he wanted to do it himself. Um, but I I was working for him for about two years. So, all of these starting with Stop Filming Strangers all the way up to uh the last one I designed for him was this tiger one. So, and you can see how um he was getting 8 million, 11 million views, 10 million views back then and now he's he's only getting two and three million. So, I don't know, maybe he'll have to give me a call someday. But, uh I'm not going to take all the credit, but >> so all of these are using um a lot of combinations of again real assets, but then also AI assets. So, for example, here, this one, my quest to find the weirdest '90s toy ever. That one is a real face. Um, but the arm is AI and then the whole background is AI, but it's based on real objects. And so you can use um as you know, reference images basically to put virtually into your scene. And again, I'll show you how I do that uh in a minute. But so it's just a real great combination of kind of everything. I think this one I used only vintage technology for a week. That one um doesn't really have any AI generation, but it has a ton of generative fill going on in that one. So, um I don't know. Let's see what else. Uh Max Foch, I made this one here. I think if I type laws, you'll see it. This one here, not that one, but this one. I broke silly laws in front of the police. Uh over 8 million views on that one. that is based on a real photo, but using generative fill in Photoshop, we were able to put the beer uh in the mug there. I made the cow bigger. We move things around. So, using AI to manipulate the image is really the best way to use it in a completely undetected way. It's a very, very powerful tool that allows you to do things that you couldn't do otherwise, especially if you don't have any graphic design experience like myself. Um, and yeah, it's it's just an amazing amazing tool. So, that's just a a quick kind of summary of just some of the clients I've worked with and some of the, you know, just a few of the little thumbnails that I've done. But, um, yeah. Do you have any uh thoughts on that so far?
Yes, I love this. This is great. Um, are you able to go back to the gentleman you were on before, >> Curtis? and scroll, Curtis, and scroll down to a video that you you didn't do the thumbnail on and kind of show us why that thumbnail might not have done as well as some other thumbnails, but so we can get an idea cuz I'm seeing like there's a lot of stuff going on. They're quite bright. They stand out. Be nice to kind of go, oh, this one's not so good because of XYZ, you know, and just break it down a little.
So Curtis has Curtis has been making content like let's switch to oldest. So, he's been making consistent content for 11 years. So, he's like he's a really consistent creator and he's got a really dedicated fan base. So, he could post a thumbnail of uh paint drying and it probably would get, you know, a million views in a couple of days. Um, so he he is like an amazing amazing talented creator with a really dedicated following. But to critique like here's a good example, this Barbie versus Bratz thumbnail here. um that uh I I made um a thumbnail that was also called Barbie versus Bratz for him. It was like a follow-up to this one. And so you can really see uh a difference between the two. So one of the things here is uh that outline around his, you know, around him is kind of a dated look. The exposure is all over the place. Um, and the detail, like the text is too small in certain areas. Um, and the composition though it's pretty good is uh not not it doesn't have any like leading lines. It feels very thrown together and very amateur. But I will say this style of creator which is called the commentary niche on YouTube. They kind of live in that DIY kind of thrown together world. And so when I work the majority of my clients uh for a long time were like this. they were in the commentary niche and you kind of want it to have a certain level of crappiness intentionally. And so I'm never making it look like a real photo. I'm always making it feel handmade, but there's a level of finesse and uh elevation to it that kind of make sure that all the fundamentals are in place, but still keeps that DIY aesthetic. So I'm going to go up to my Barbies and Bratz uh thumbnail. So there you go. So, here's my version of the same thumbnail, Barbie versus Bratz 3.
You can see that there. Um, so again, we're taking a lot of the same elements in terms of the composition and the layout, but I simplified the posters. There was text all over those posters, but I cleaned it up using Gen Fill. Um, I used a color that, you know, kind of matched. So, we've got the same pinks, we got the same purples, the same blues. So, there's like a real color balance to it. The exposure on Curtis's face is clear and straightforward. And then I also took those uh movie posters and pointed them in to his head to give it a 3D effect so that those leading lines is what it's called are going into his face. And so again, when you compare it to the original one, which was down here, you can automatically see kind of the difference between the two, right?
So yeah, absolutely. That's a good example of of uh of that, I guess. Um I actually do a whole breakdown on that on my other channel. Um, let's see. I did a whole like breakdown. You can see my process on that particular thumbnail. So here's um yeah. So, here's the actually this is a I can just walk you guys through this. I'll play the video and you can see that's the original asset for that thumbnail, the Barbies versus Bratz thumbnail. You can see when I do the generative fill um or the uh subject selection tool in Photoshop, it's using AI to select Curtis. But you can see some little uh issues with the hair. And so um using Jin Fill, we clean that up. Let's see. I think I do it around Yeah, right here. So check this out. Right here. I take my little mouse and I get to it in a second. Sorry. Uh, I take my little mouse and you can see how the hair is like it adds the edges to the hair using AI. So that's like I'm using AI constantly for cleanup purposes. So like it totally fixes edges and I don't know what I'm doing here with uh the video is a little too fast edited I think. But um then you can see too my process with uh the leading lines like that's what I was starting with. You can see how I'm using Ginfill to clean up the text on the car. So that you can see that if I pause it. So that's what the Barbie poster looked like before on the left. And then using Jin fill, I was able to extend out the car there and remove the text, which cleans it up. You don't want to have any unnecessary text or anything that doesn't need to be there.
Yeah, I think I did the same thing for the Bratz thing. I got rid of the TM logo there and everything. Um, yeah. And then I also like uh I did a little video edit where at the intro of his video, this was my idea to Curtis, is like what if you like said your intro and it matched the thumbnail like literally. So this is a example of me animating a little uh segment that's actually the intro of the video. So it's like the thumbnail comes to life. Um, and I reveal that to Curtis and he really liked it. So there he is on FaceTime. I was showing him the thumbnail, so you could check out my video later. But, um, anyways, I I think I was rambling a little too much there. But, um, you know, it's it's a really powerful tool. Gen fill is what it's called in Photoshop. Generative fill. And it's just gotten better and better. I think I made this video like about a little over a year ago. So, it's it's continually getting better. And it's a very powerful tool because you can isolate certain areas of your image, tell it, you know, type in things like fix the hair, remove this object, whatever. And you can directly communicate to the AI. And what I love about it is it's like object aware or like camera lens and color science aware is what I say. Meaning like if I were to generate an object using Midjourney or Google Nano Banana and then bring it into my composition, it's not going to know the color and the way that the light was positioned in that photo or whatever or the way that the camera looks. But when I do it inside of Photoshop and I isolate certain areas and say remove this object or extend this thing, it's actually taking into account the color, the lens that the camera was using, the bokeh, like which is the blurry background behind a camera. Um, and it really really looks like it's in camera basically instead of adding something in from a separate tool. So anyways, it's a very very powerful tool and I think anybody who hasn't played around with Photoshop should give it a try. That's why we started Thumbnail Academy. Me and my business partners uh Diltoma and Antioch. Uh, they are some of the biggest uh thumbnail designers in the world. They work for creators such as Dude Perfect and Faze Rug and many others. We have over 10 billion views of experience between the three of us. Um, and we're teaching everything we know uh with these tools. And I haven't even touched the surface on Freepick, which is our main tool. But check out Thumbnail Academy if you want to learn more. It's where we teach everything because now there's really no um like barrier of entry. It's really gotten to a point where you just need to learn how to use these tools and learn why. That's the big one is like understand the why to get the click. Right now, AI is not replacing the YouTube strategy knowledge that's required to be good at YouTube, but it is replacing the work of a lot of tedious hours that used to be spent designing things in Photoshop. So, like you said to me earlier before we started recording, like it would take hours to just cut somebody out of of a frame and then use that as an asset. Now, you can do that in two seconds using subject select in Photoshop. So, anyways,
Absolutely. Yeah, I totally agree. Like I sometimes I would join or combine three images for a YouTube thumbnail and it would take me over 16 hours because I'd be cutting bits out and recoloring things and all of that sort of stuff. So, I love all of this. Are you able to dive in and show us some of your workflow? And even I'd love at some point, I don't know whether you want to show us the tools first, but just breaking down if someone says to you, I need a thumbnail. What's your process? Like how do you come up with what goes on the thumbnail? Where's the idea come from? like I'd love to hear both about the tools and the ideas.
Okay. Um, let me find something here. I'm just Okay. So, it always starts with the overall concept from the creator. Um, and I use Figma to kind of build out a a a board for the creators I work with. So, this guy that I work with, his name is Jayl Klouse. He's a YouTube uh and creator educator kind of guy and he's interviewing >> your phone. >> What does Figma do? I've used it before, but just it's not that well known. What does Figma do?
So, I I don't know too much about it either. I think that it's mostly used by web designers and coders and fancy people, but I just use it as like it's basically an infinity whiteboard. It's like a virtual whiteboard, corkboard, scrapbook, and I think it works really well. Like this particular type of tool that's built in, it's totally free to use this. This is like uh what's it called? It's called a uh a jam board, I think is what they call it. And it allows you to just simply take images. Like I'll take like a screenshot of something like I'll just like here I'll use my Mac. I'm using a the screenshot tool to do that. I just took a screenshot. And now I can just drag it in to I don't know if you could see that because I'm not sharing my entire window, but basically I just drag that image into Figma. I can scale it up. It's like it's a vectorbased tool. So I can scale things up. There's like arrows. So I can, you know, if I'm trying to communicate this goes to this and I want to add some text, I can say, you know, poopy pe or whatever, right? So, um, it's a really powerful way to just kind of have a virtual whiteboard, I guess, >> for lack of a better term. So, um, so anyways, so this creator that I was working with, he interviewed this lady who is a lawyer specifically for for creators, for YouTubers. So, she helps YouTubers avoid getting in trouble with fair use or if if a creator gets sued for some reason. She helps uh she helps them through that the lawsuit and stuff like that. So, but she was just on this is a screenshot from her Zoom call. She was just on a webcam. She didn't have a nice camera. You know, it's not terrible. Like obviously the lighting isn't actually that bad, but um, you know, he was a little my the guy I was working with who interviewed her, he was a little concerned. He's like all I have is just this shot like we might need to use a proper head shot from her LinkedIn profile or whatever. I was like oh no we'll be we'll be fine. AI will solve this problem for us. So um so this is again the shot of the lady that he interviewed. Um, this purple background is the background that we've consistently used for his studio. It's a virtual background that I think a designer designed for him back in the day. Like I think this was made in Blender like a couple years ago. But let's just imagine that this is your like real studio or backdrop of your set. So using a tool called Freepick, which I'll show you how we do it. All I had to do is just input this image of this lady and then input this image of the background. And then my prompt was something like take this reference image of this lady and put her in the set uh uh off center because I don't want her in the middle. I want her kind of off to the side with the podcast microphone looking like she's midsentence or something like that, you know. And I said, "Make sure it's photoreal, etc." And then we were able to create, you can kind of see it was already peeking out. We were able to create a couple of um versions of this where not only did it put her in the actual environment, but it totally like fixed the resolution problems that we were concerned about. It like upscaled it quite a bit. You can see like the quality when I zoom in on Figma, you can see it kind of goes into this weird pixy mode. So I if I zoom in more than 400% I guess it it doesn't look that good on that's just on Figma. Uh, let me see if I actually scale that up. There you go. So you can see how much detail it added to that compared to this low res zoom call. Right.
Yeah. So, basically reconstructed her face for me. Um, and that's because, uh, I was able to build a reference image on this shot and then it just used AI to fill in the gaps. And so I did that a couple times and then we landed on I guess I think I landed on this one here >> that I was just showing as my main. Yep. But here's the thing. Here's where it's it's crucial as a thumbnail designer. You don't want to just run with the first generation because there's a couple things I don't like about this. Like I don't like how the bookcase looks kind of weird. Doesn't look very realistic. I also don't understand why there's a gavvel right there. Like a a judge. Is that what it's called? A gavvel? Like the hammer thing? Um, that's that's dumb. Um, and then uh she's not off center. I needed her off center because I was going to put text on the other side. Um, and then I don't like the shirt in the background, right? And so using Photoshop, >> I was able to use subject selection to select her. And I could um should I just uh show you how I would do that? Actually, let me just >> Yeah. >> Can I share my Photoshop screen? >> Yeah. >> Okay, hold on. Hold on everybody. >> Let me share my screen with Photoshop instead. I'll just do my entire screen. How's that? Okay. Okay. That's another thumbnail I can show you later. Um, please hold. Please hold. And with a lot of the people that we've worked with with uh inside of Thumbnail Academy, I've noticed that most people are using Canva for thumbnails who are beginners. Um, stop stop using Canva. It sucks. It's terrible. It's trash. Stop it. >> It's not professional enough. And, uh, it doesn't have the right tools, believe me. Uh, they, in fact, Canva themselves know that. That's why they acquired uh I think they acquired Affinity. So they now offer like a pro tool. I think it's called Affinity. And um the pricing structure is the exact same as Photoshop. Like Adobe Photoshop is uh I think I pay $15 a month for the photo only bundle and it's got all the tools you would need and it feels overwhelming if you've never used it before. That's where Thumbnail Academy comes in. That's where tutorials on YouTube come in. It's not, you don't have to know how to use every tool in Photoshop to uh really take advantage of it. It's so much faster. It's so much cleaner. Trust me. Do yourself a favor and uh start learning Photoshop. It's worth it. Photoshop is like the most popular uh pro tool, procreative tool in the world. And it's um, you know, it's Adobe's number one product. They make a lot of products and Photoshop by far is their number one mostus uh app. So, learn it. It's powerful. Trust me. Um, okay. So, we're going to paste. We're going to put that uh that thing of that lady. Okay. So, let me Hold on. Uh where is it? Hold on. Sorry. I'm just going to screenshot it because I'm not figuring out how to save it. So, I'm just going to drop it in. This is using my overlays um that you can download on thumbnailacademy.comresources. And what it does is it shows me my composition with the rule of thirds, which is a whole another can of worms that maybe I won't get into today, but basically it it's a uh a rule that like our eyes are just so naturally designed to kind of enjoy and and have balance around this. Uh, you can almost imagine your image like it's on a scale and if you have a lot of weight on one side like we have here right now like there's weight on this side with her face. So we need to balance that out by adding weight on the other side with in our case we use text to to add that weight. Um, so using this is really powerful. And then we also have both mobile and desktop versions where you can see where the time code is going to live uh on YouTube because anytime you're on YouTube, that time code is always covering up. And on mobile, it's huge. So you cannot ever put text down there on that bottom right. It just gets totally covered up. So check this out. So forget about the composition for a second. So, I'm going to frame her the way I want, which is about there. So, I don't like the gavvel, and I need more room off to the left of that bookcase. So, I'm just simply going to select this rectangular marquee tool, or you could use the lasso tool. But since a square selection tool is perfect for this, I'm just going to use that. I'm just going to drag it over like this. So, I'm selecting this empty gap. I hit generate generative fill. I don't have to type anything at all. And I hit generate and it just uses its own intelligence. It's Adobe intelligence to fill in that area. And I promise you, it's going to look halfway decent. If it's not, you can always there's no penalty for Wow. >> trying it multiple times. But yeah, there you go. There's one, two, three options there. >> Uh I can run I could run this as many times as I want and it doesn't matter. So, um, now let's remove that gavvel cuz that's stupid. So, I'm going to select it like that. >> And that's the thing is like >> that gap fill thing is insane. Like that saved me hours when I was younger of like when I was doing a lot of stuff by hand in Photoshop. I would have to try and find backgrounds that were like fit everywhere and have all the things in them and I'd be cutting and copying and pasting and, you know, fading stuff like Yeah. like thing. >> Yeah. I mean, and that's the thing is it used to take me before all these tools came out, uh, it would take me four to six hours to make one really pro thumbnail. Now, we're able to get it done in like 30 minutes, you know, which means we can just be that much more efficient. Um, and I love it because the more thumbnails I make, the more I learn about different niches because I have the benefit of being on the other side of this where I'm working with dozens of creators and dozens of clients or uh, you know, a lot of corporate companies too I I work for in their channels and stuff. And so I get to see different niches on YouTube and learn what works and what doesn't work. And so it's really really fun. And these tools allow us to be just so much faster. So but yeah, so I'm going to remove the gavvel. So, I just switched over to the uh lasso tool. I kind of like using this one. It's called the polygonal lasso tool. It kind of you can see how it's a uh it allows me to just with one click create these kind of shapes and stuff. I like I just click once it creates a point and then I can drag it however I want wherever. It's it's just more convenient than the normal tool where I'm literally drawing like I'm drawing like it's a paintbrush now, you know, >> and that can get a little messy. So, because I'm just using a mouse, I'm not using like a an iPad or something. So, so I'll select this thing. I'm going to hit generative fill. And then I'm just going to say remove. And so, again, it's it's still just that simple. And um it'll it should remove that object. And it does it um, you know, fairly quickly. It's all dependent on your uh your internet speed because it is going up to the cloud to do this. So, yeah. So, there you go. So, there's one, two, three options. Wow. >> So there you go. So again, I'm going to toggle on my rule of thirds grid. Now again, I don't like this little thing over here in the corner. So you see how I'm sort of going over her jacket because I want it to remove this object but still keep the edge of the jacket intact. So I'm going to I kind of had the selection go into her jacket a bit. So I'm just going to again say remove. and it's able to understand enough of what I'm saying. And I'm not even trying to really make this that good of a selection. It's, you know, it's pretty rough because it's it's able to just do it pretty well. So, there you go. So, it's gone, >> right? So, now we've got a nice clean base here. But again, I don't like the background. It's too It looks too fake. Um, but that's fixable with bokeh, which is the the lens blur effect. Um, also I do want to have the freedom to move her around. So, I'll show you how I do that. So, there's a hotkey um that we teach in Thumbnail Academy. It's shift commandoption E. Okay. So, it's a it's a lot of buttons, but what that just did is it created like a mixdown of that layer. So now I have, if I turn off all these other layers, you'll see it's just one layer now, which is nice. So I'm going to select, you can see the select subject tool like that. Boom. And you can see how quickly it selected all the edges so beautifully, so perfectly. >> Yeah. >> Um, her hair is not perfect, but that's okay because she It's not like we're going to cut and paste her on a different background. We're going to cut and paste her on this exact same background. So, if there's little edges here and there um from the previous background, it's fine. It it'll kind of blend into itself. So, I'm going to basically hit um command C. I'm on a Mac, which is copy. So, whatever your copy is. And then I hit commandV, which is paste. You can see there I now have her isolated. >> Yeah. >> But now I'm going to do it again. So, I'm going to hit subject select. I go up to help because I don't know where this is. I I never took the time to learn where this is, but I know it's called expand. Where is it? Modify, expand. Okay. I always forget. And that's where the help tool comes in is like what what's that tool? I know it's here, but I don't know where to find it. You just type it and it'll find it for you. So, there's expand. >> I usually expand by seven or 10 pixels. And you'll see here, watch. It takes that selection and it expands it out. So, if I zoom in, you'll see now the uh the selection has gone a little bit outside of the bounds of her hair. You can see how it's like a couple pixels outside of her jacket. That's what you want to remove her from the background. If it if I didn't do that, I I just hit undo. >> There's like little edges of her jacket that aren't selected. There's little edges of her hair that aren't selected. And if I said remove this object, those little slivers of her jacket and those little slivers of her hair would still be there. So it'd be like everything vanished and then just like the little outline of her body is still there, you know? So >> yeah. >> Um in fact, I noticed that her hair was still not fully um was not fully removed there. So I'm going to expand even more. >> So I'm just going to go to I'll do 20. There you go. Perfect. So, that that cleaned it up. I noticed when I was zooming in that some of her hair was actually still uh in fact, there you go. Some of her hair right there is even still getting covered up a bit. So, there you go. So, now the selection is outside of her for the most part. I could have done that by hand a little bit more crudely, but I just like doing this. I hit generate fill, hit remove, and now it's going to give me a nice clean background, which is really crucial because now I I took one flat image that was just a JPEG, you know, that was was generated in AI in the first place, and now I've been able to expand the set in the direction that I wanted it. There you go. See? So, she's gone now. >> Um, and now I've got a clean asset to to work with. And there she is. Right. as an additional layer. >> But now as a human, I can decide where I want to put her, right? >> Y >> and that's crucial. And that's I think these steps that I just took >> are what is needed to really, you know, see yourself more as an artist and and have more intentionality because AI certainly does the thing really well, but there's just little things that it misses and there it it basically just misses the mark often. um it's just a little off on the composition or >> I don't know. I can't say it any other way other than um it doesn't have taste because it's not a human. So it just it doesn't think the way we think artistically. >> I you know I believe in God and I believe God created us in his image to be creative people. And AI doesn't have that. AI was not made in God's image. So it will just simply never have the ability to truly create. It'll have the illusion of creation but it will not actually be creative like you. So I think if anything that's hopefully a hopeful thought is like at the end of the day as humans we are creatures and beings with a soul and a li and purpose whereas AI is just simulating and emulating souls with purposes. It doesn't actually have a soul or a purpose obviously. So uh there's my philosophical rant. Um, so now Uh, now I'm going to do that mixdown again, which is shift commandoption E on a Mac. And again, what it did is it created an image blending all those layers. Now, check this out. I'm going to go into filter, camera raw, and I know this is overwhelming. This is a fire hose of information. There's no way you guys are going to remember all these steps. Again, that's why we built Thumbnail Academy. It's a full multi-hour course beginning to end to teach you everything you need to know about this. you can always refer back to it um and whatnot. But if I scroll down here, this is another AI tool uh built into Photoshop called lens blur. And it's using AI to determine what object should be in focus and what object should be out of focus. And it applies a gradual blur to the background. Unlike how we used to do it back in the day where you just do a gajon blur to the whole background, everything is blurry here. Actually, let me show you what that looks like. So, I'm going to mix down my background. And then there she is. So, I'm going to show you. This is how we used to do a blur back in the day. It's a gajian blur. This is like how Zoom and Google does a blur on a live stream. Are you doing it right now? Are you doing that effect? >> No, I No, I don't like that effect at all. >> Exactly. So, it's just like a general blur. And it's not terrible, but it does not look anything like a real lens like this. By the way, you know, I'm using a real camera right now on my little webcam here. So, you can see this background behind me is out of focus, but it gradually goes out of focus. So, that's what we're going to try to simulate is the look of a real lens. And this tool is a total AI tool creating depth. So, I just hit apply and it'll literally apply a depth map. If I turn the visualize depth tool, you'll see it was able to determine that she is the foreground element. This bookcase on this edge is slightly closer than the bookcase back here because it was able to use AI to determine, okay, this is a bookcase going with leading lines going back. >> And you can see like where the this there's not a ton of uh rich detail going on here. It's a pretty straightforward image. Um, but you can basically change the focus like you would on a camera lens. Isn't that cool? >> I honestly can't even believe that Photoshop lets you do all this stuff now. Like I really felt like >> I know, right? >> Photoshop was either getting too complicated or left behind by other AI tools. So, the fact that like Photoshop has this much um ease and functionality, like you're doing this so fast and like literally we were not doing this this fast, you know, eight years ago or five years ago. >> Exactly. >> So, like the fact that Photoshop is doing all this now is blowing my mind. >> Yeah. Oh, yeah. For sure. So, I'm I'm fixing some of the edges here. You see how when I >> It It didn't get that right. So, I just go to refinement, I hit focus, and then now I can just kind of paint I basically just paint the focus back in. >> Wow. >> To that little edge that was kind of weird. and wacky, right? So, I can kind of >> And again, people are never going to look at your thumbnail this close um because thumbnails are mostly viewed on mobile and it, you know, also they're viewed on TVs for sure, but um they're not really pixel peeping your uh images, but you can see like pieces of the hair are a little weird. Um, so I go in there and kind of refine that, but it doesn't have to be perfect. And you can see it applied a really natural blur. And what that did is two things. It made it look that much more realistic, that much more like a real photo, which again, like I'm always trying to make an AI image look as much as uh as much, you know, real >> what's I don't know how I should phrase that, but you're trying to make it look as real as possible. There you go. That's how I should have said it. >> Um, and but then it's also intentional. Like there's actual real intentionality behind how I'm doing this, why I'm doing it. you know, the way that I scooted her over is the way I wanted it. I was able to expand it the way I wanted it. So, I'm I'm forcing the tool to do what I want, not just going with whatever the robot is giving me. So, um, but obviously, again, this whole process wouldn't even be possible without it. So, I'm not I'm I'm definitely not discounting the power of the tool. Um, so there you go. So, yeah. So, then I would just say like, you know, >> get your life right? You know, whatever whatever you want to say with your with your text. >> I uh let's see. I don't remember what he uses. I forget what font he uses. That's a whole another discussion. Uh, you know what? I think it might be >> uh h No, I don't remember. I don't remember which one he uses. I'm just going to use Helvetica. Uh, it's one of my favorites. So, um, you know, so we could switch that to maybe a bold a bold font. >> Um, and yeah, you kind of get the idea there. And obviously, I can go further and further if I wanted to. I think I do like a little stroke um underneath certain text. So, I'll like add a add something to it to give it a pop, you know, like that. Oh, yeah. >> I'll I'll paint a little bit of darkness behind it. So, I just switched over to a black paintbrush and I'm going to create a layer and I'm going to just manually Well, I don't want that. I don't want it to be that hard. So, I'll soften that up. >> Yeah. >> Um, and then I'll just paint paint some shadow in there. Like, it doesn't have to be perfect. Um, and then you just lower the opacity down and it kind of maybe change the blending mode to like linear or something. And uh, yeah, just kind of helps it pop a little bit. Um, yeah, so you know, you could that's a whole another thing, but you get the idea. That's I think the final thumbnail uh is in here somewhere. YouTube won't help you. No, that's not it. It's um this guy has at J somehow. Jay somehow got this handle. So yeah, so creators are losing creators are losing their channels. Yeah, >> it's got 100,000 views. You know, this channel, I think he has like 100,000 subs. So, this was a one of 10 performer for him. And that's the final thumbnail that we went with. Looks like I used a different pose actually uh in the final, but it's the same exact process. So, >> um, yeah. So, there you go. There there's that little example.
So, I'd love to know two things because you did mention a thing called pics or something at some point. I'd love to know about that. But >> free pick, I'd love to know about that and what that does. But also, I'd love to know, or we can come back to this, if someone was trying to design a thumbnail, what are the three main things that they should be looking at trying to get like pretty good?
Yeah. So, um, there's a lot of a lot of uh things there, but we actually have four things. We call it the four C's. Um, and that's what we teach. We uh I say we because I have business partners D and A. uh with Thumbnail Academy. We actually were uh fortunate enough to speak at VidSummit this year. Um, I spoke last year, but the three of us spoke this year. So, let's see. Where is it? I'm trying to pull up a slide. I had a slide uh with the four C's, but uh hold on, hold on. Ah, dang it. Um, yeah, so composition is one of the four C's. Color theory. Um, curiosity and gosh, I don't remember what the other one was. That's why I'm trying to pull it up. Um, I don't remember my own teachings. >> Here we go. Thumbnail talk. >> How to create winning thumbnails. Here we go. >> Oh, is it sharing? >> Yeah. Yep. >> Cool. So, yeah. Composition, color, clean assets, and curiosity. Okay. Yeah, I forgot about the clean assets. So, we kind of already talked about composition. Um, >> yeah. >> You know, you got to make sure that your your image is balanced. Well, there's something to, again, as a Christian, I believe that God created us to desire order because he's an ordered God. Like, this universe is too perfect and fine-tuned for it to be an accident. And so, we are attracted to things with order. uh when chaos happens it just uh it gives us stress anxiety and you don't want that. Um, so yeah, having order and balance is important because it is attractive to put it simply but also watch out for the time code like we talked about in the bottom right corner. Uh color balance. So understanding color contrast,
color balance. Uh, this is the color wheel. So, learning that, I learned this when I was in college. I only went to college for a year and a half. I was going to go for graphic design. Um, and we learned the color wheel, which is applicable to painting, to filmmaking, photography. All those tools that you would learn in any graphic design class or photography class are all applicable to making good thumbnails, too.
You can see here some of these are just some of the thumbnails that the three of us have designed. Um, out of these, I designed this one here, the Killian Murphy one. That one's mine. And the other ones are Dylan Ant. But you can see how we use, um, like the red contrasting against the blue background with the green grass.
You know, in this case, I've got the Batman logo, uh, or the Batman bats on the left with the Inception buildings on the right with the atomic bomb in the background, and then Killian Murphy. So, it's a, it was a film. It was a YouTube video about his career. And so, it's kind of showing all three of those movies in one. And then you have that color contrast going on. Color theory, red, uh, equals, you know, it makes you feel more, uh, danger, urgency. Blue, intelligence, and calm. So, there's different feelings that you have when you see certain colors.
Yeah.
Clean assets. This was, I think I showed this one earlier, one of the thumbnails I made for Curtis. Um, that's the real picture that I was given of actually Curtis. Like, that's really him. He had this device, and that's him in front of, uh, a white wall. And so, because the lighting was so good, because the white wall was there, I was able to very easily cut him out using the subject select tool that you saw. And then all these other assets were from, um, stock photos. This was before I was using AI to generate my assets, but now we would use AI to generate clean assets in the first place. It's actually better to basically create the individual elements and then compile it yourself in Photoshop sometimes.
Because again, that human touch is needed to give a sense of order, uh, and intentionality.
So, if you imagine again, these are all real photos.
Yeah.
Um, but you could totally replace, like, those two subjects on the bottom, even Curtis, with an AI-generated image of, like, generate an image of a lady wearing a green hoodie looking confused on a white background. And then you just do that with all of your elements, and then you pull it into Photoshop, and then you can manipulate it by hand, and it has that human touch. So, that's, that's what I'm doing a lot now.
Um,
I love that.
Uh, yeah. Uh, let's see. You can see how, uh, you can use an iPhone to create, uh, really great assets as well. It doesn't matter. You don't have to have a nice camera. Um, this is all talking about this is when we went to, uh, Mr. Beast headquarters. We got to see the process behind this thumbnail. That's us right there in the Beast compound, and we got to see how they made this thumbnail. Got 100 million views in a couple days.
Um, yeah.
So, curiosity gap. Um, you basically want to have a thumbnail and title that creates a question in somebody's mind that they never thought of. So, here, here's a really good one. How bad is McDonald's really?
It sets up a question that can only be answered by your video, you know? So, it really, like, sparks some sort of, like, "Oh, yeah, well, maybe you never really thought about that before." But it's like, "Okay, well, I'm expecting that he's," and Brian Johnson is, is well known for, he's that billionaire guy who, like, is optimizing his body for health right now. So, if you recognize him with that title, you know, "Okay, he's going to go do a deep dive on this, and I'll, I'll get the answer to that question."
Uh, you can do shock factor, you know, walking on water prank, um, all sorts of stuff. So, yeah, um, good clickbait is important. Like, this is the actual object that this YouTuber created. And then this is what the thumbnail looked like. It's still the exact same object, you know, with, you can see how it really is those same blowers that he installed on the hovercraft, but they exaggerated a bit to kind of help, you know, get the click, get the interest. You want people to see and understand it immediately, but don't mislead them. Like, this isn't a misleading thumbnail because he makes a really cool hovercraft, but it doesn't float like that, you know? So, understanding that.
Um, and then, yeah.
Yeah.
Anyways, what was the Freepick thing you mentioned? Because I'd love to know about that, too.
Just, okay, so Freepick.
Yeah, what are we doing? Yeah. So, Freepick is, um, there's other tools like this out there like, I think Higsfield is a really popular one, too, but all like me, Dylan, Ant use Freepick. We only use Freepick. Um, it's a really, really powerful tool to kind of compile all the things together. So, I'll show you an example. I just made a thumbnail for a, a channel that I work with called Black Rifle Coffee. They make, um, they're a bunch of military guys who, uh, they're veterans, and it's a very American, uh, a very American, uh, company. You're from Australia. So, uh, you guys don't have guns like we do. So, um, these guys basically took, like, this is a full AI image that I generated. Um, let me recreate it and show you how I did this. So, basically, like, this tool is basically like your Midjourney or your Leonardo AI sort of thing.
That's right. Yeah.
Why do you use that over a Leonardo or a Midjourney? Because everybody talks about Midjourney, but I haven't heard that much about this tool.
I'll show you why in a second. Yeah. So, here's the, um, here's the video. So, if you click on the video, you'll see, like, these guys are maniacs, right? They installed, um, insane weapons on these vehicles, and they're just being silly. And, um, you know, this is a billion-dollar company, Black Rifle Coffee. They, they're in every store in America, really popular coffee brand, but it's like owned by all these military guys, and they donate a significant portion of their revenue to to veterans and, um, yeah, so they're, they're pretty awesome, you know, badass, if you will, right? So, um, you can see though, like, all I need to do to make my. So, I, I already had an idea in my head. I want all three of these vehicles. They retrofitted a Tesla. They retrofitted a Prius with a, with a gun. Where's that? They put a gun on a Prius. They put a gun on the Bronco. So, all I had to do was take screenshots like that of each thing. So, here's the Prius. Here's the gun, right? And I'll, I just for sake of, uh, this, you know, course, I'll, uh, I'll, uh, save you. I don't have to do all that, but let me reuse this, basically.
Um, so, I'm going to recreate it. So, you can see here over on the left, I have all those screenshots. So, I don't know if you can see it really.
Yeah. Yeah.
It's kind of small, but you can see there's the screenshot. Image two is a screenshot of the Prius. Image three is the Prius. Image one is the Bronco. Uh, there's the guns from the Tesla. And then there's the shot of the Tesla. And so, I basically was able to use a combination of, and I'm, I'm usually prompting things with AI as well. So, let me go to the prompts. I think. No, that was the other one I did. Ah,
Okay. So, I'm making.
Yeah. Yeah. Uh, is it? Is that it? No, that's not it. Hold on. Hold on. No, you know what? I didn't use a prompt for this one. No, I just wrote this one by hand. Okay. I just tell it I need a thumbnail photo of these three vehicles with guns on them and the Black Rifle logo, which I've included at image seven. Uh, and so I basically reference each reference file as I say it. This app allows you to build characters as well. So I can add any character that I want inside of here. These are all just the creators I've worked with. Um, and so if I click into that guy who's on top of the car, his name's Matt. Where is he? He's down here somewhere. There he is. There's Matt. So, I was able to to say, you know, um, image one is a blue Bronco with a gun and and a man who is on top firing it. And I'll say at Matt, right? So, I'm telling it that this is the character that you've already built your model on. Uh, image six is a Tesla. Model X with guns that look like at image five at image four. So again, I'm referencing the the screenshots that I took from the video of the guns on the Tesla. Image six is a Tesla, uh, or what is it? Image two and image three is a Prius with a big gun on it.
Right. So again, I'm just very clearly communicating and not overdoing it, just telling it what it is. I need all three to be lined up shooting towards the camera, make it look epic, highlight the guns firing, and also use the logo and make it look as close to the images I provided. So that's all I said. I'm able to select 16 by 9. I believe the winning thumbnail that I did for that one is this one, I think. And I was using the newest model of Nano Banana. So, I guess that one is, yeah, Google Nano Banana 2. So, this was the one that I chose.
Wow.
And that's based on my prompt. So, you can see, like, what I love about it is it was using all those reference files from the screenshots to inform this design. Like, it's the same green Tesla that you see in the video. It's the same blue Ford that you see in the video because I built a character model on that talent. The the guy, his name is Matt.
The guy shooting the gun looks a lot like Matt.
Yeah.
And then of course, you've got the logo with the Prius with the gun on top. But again, you can see, like, I don't want the logo up here. That's messy. There's some sort of weird thing going on on the the door wings there. And then these bullets. I don't like these bullets flying here. It looks very unintentional. And so then I, uh, pull that into Photoshop and do what I just did, which is.
Yeah.
Clean everything up, right?
And so if I disable all the layers, like, this is what we just saw. Again, I'm using my overlay to rebalance it.
Yeah.
To give it some intentionality. I'm going to extend the sky up. I'm going to remove the logo. I'm removing the the weird gun trails that are there using Genfill. Remove, remove, um, the doors have these weird things on them, and I remove those. And so there's that clean image. And then I do, like, my color, uh, my color pop and saturation contrast pop. And there you go. So, there's the final thumbnail.
Yes, it's AI, but I'm manually fixing all of the things that make it sloppy, you know?
That is amazing.
Yeah. So, this video is doing really well so far on the channel. Obviously, the, the audience loves it. Um, because they're a military audience, and it's an American, you know, exactly how I want to see my coffee club profits used. Uh, Black Rifle, Black Rifle Coffee's motto must be "Because We F***ing Can," and I'm here for it. So, uh, you know.
Came from Norway. Hey, we got some Europeans that are all about the guns, right? So.
Yeah. So, the people love it. The video is doing well. Um, and yeah, and because these are the vehicles in the video, I you must make sure that you're not misleading the viewer. So, that's why I go to all this trouble to make sure that the cars on the thumbnail are accurate to what's in the video. Cuz I could just like generate a cool Tesla with a gun on it, but it's never going to look exactly like the one in the video. And even if it's a red Tesla on the thumbnail, but then in the video it's green, that's misleading. That's not telling the truth of what's in the video. And even though somebody might not realize it, like, it might cause them to think, "Well, you know, I don't really want to watch this anymore. They weren't, they were kind of misleading me, you know, or whatever." So, I try my best to like use those references from the video as to inform the generation that is in there. The reason that I use Freepick is because I have access to all these different models. Uh, right now, Google Nano Banana 2 just came out. It's pretty good. I don't find it that much better than the than Nano Banana Pro. Um, you've got Seedream, all the different models here. Um, but the reason I use this is because of this way that I can build the character models for all my YouTubers that I work with. Um, which, like, if I go in here to, like, let's say, I mean, here's me. So, I go in here, I hit edit. Uh, I just, you only need to upload like three images of your creator in different poses, and then it creates a version of that. Like, it creates a, a digital double, basically. I called this t- this character, uh, David Christmas, because it was a Christmas video that I made. So, let me clear everything. I'll do a faster model because Nano Banana is slow. Uh, I'll do Seedream. So, I'll do David Christmas, um, you know, holding a giant lightsaber in the middle of the forest, right? Just for the heck of it.
And it's going to generate me in that same outfit with probably headphones or in a weird way. And yeah. So, using the character model built into Freepick, it allows me to take images from the video again so that we're not misleading the viewer. Um,
Yeah.
It looks like the creator in the video in the generations. So, that didn't work.
That did not work at all. That was trash. Uh, hold on. Let's do Nano. I, I typically use Nano Banana for everything. That does not look like me at all. That's absolutely terrible. I am still getting used to that's the new Cadream. Uh, I've been using Cadream. Creree 4.5 is better at, uh, character consistency. And how do I know that? I just experiment. I just, I've used all these models. I experiment with it. And so I kind of know, like, "Okay, Nano Banana Pro looks the most realistic. Cream 4.5 is pretty consistent with character consistency. It looks like Cdream 5 did not hold character consistency at all." Um, that's not working either. What's going on?
All right. Well, you can see, like, sometimes it works. Like, this is a creator I work with, Ted. That looks like him. Like, you can see the reference images of him over here.
Yeah.
That totally looks like him. But then sometimes it just, like, wigs out and creates, like, a weird version of, like, that does not look like Ted, you know? That's a different, that's a different person. So, sometimes AI will just kind of start acting weird on you. Um, again, yeah, this is not working well. So, forget all that.
Trial and error. It's just playing with it and trial and error.
Exactly.
So, um, yeah, I mean, here's even, like, a silly little thing. This is, I'll just share with with you on this one. My son is really into making comic books, and so we are taking, um, his drawings that he drew on paper, and then I put it in Freepick and say, "Remove the paper texture, make the pencil lines much darker so it's either full white or full black."
And so it turns his little paper sketch drawing into a much, you know, sharper, cleaner black on white image, which we, we've been printing out and putting into little comic books. You know. So,
That's adorable.
I've been using that for that. But anyways, sorry.
I love that. Yeah. Well,
Freepick is powerful. I could go on for a long time, I guess. But.
Yeah.
We probably come to the end, I guess.
Yeah. I have loved this. I'm like, I, I'm like, I want to keep going, but I'm so aware of your time as well. But this has been amazing. Like, I feel like you've given us so much value in showing us how to use Photoshop and also Freepick as well, and helping us understand why we would use that. Like, being able to have those character references of actual humans is amazing. And so, like, this whole thing and just showing us what thumbnails that are good look like and what thumbnails that aren't so good look like. Like, this has been so amazing for everybody who's watching to just basically and essentially do a very deep dive into thumbnails. So, thank you so much for your time today.
Of course. Yeah, I really believe that if you can figure out your packaging when it comes to titles and thumbnails, if you're a YouTuber, that is the ticket to success, uh, on YouTube, because YouTube is a click and watch platform. It's not a video platform, which means, you know, if you don't get the click, nobody's going to watch your video, no matter how good your video is. And these AI tools have empowered me to do things faster and honestly do things in a way that I just simply couldn't do before because I don't have a traditional Photoshop graphic design background. I have an eye for things because I was a photographer and a filmmaker for so many years, but I didn't know how to execute it because Photoshop was truly very hard to use. But hopefully, you were able to see how quick and easy these tools are to use. Once you learn it, it, it certainly can be overwhelming when you see all the things I'm doing. Like, you don't have to remember everything. And again, that's why we're starting our channel called Thumbnail Bros with me, myself, myself, Dill, and Ant, uh, teaching everything that we're learning in this process, um, for free on YouTube. But then, of course, thumbnailacademy.com is where we're going to be sharing our course, and you can pick that up if you're interested to learn more and dive deep, uh, into thumbnail design. And I think anybody can do it now. It doesn't matter how skilled you are, whether you know Photoshop or not. You can learn it, you can do it. It's a lot easier than you think because AI has empowered us, and it's an amazing thing. So, thanks for having me.
Yeah, it's just, just know Canva.
Yeah, just know Canva. Don't use Canva.
Um, now, where can people find you now? Because I know you said, um, that your new channel is coming, but if people are wanting to follow you now, where can they find you?
Yeah, I'm most active on Twitter or X. So, you can follow me at DV D Altizer. It's like DVDS, DVD Altizer. I removed the vowels in my name, David, right? So, um, alt ti zer. Check me out on Twitter or, uh, or Instagram, DV D Altizer. And feel free to shoot me a message, say hi, tell me that you, uh, watched this here. And, yeah, be happy to have a conversation, help you out if you have any questions. So, yeah, follow me on Twitter or Instagram.
Love that. Thank you so much for being here. This has been amazing.
Thanks for having me.