[00:01] thumbnails in Photoshop or Canva, you know how long it takes to get one that found a tool that makes professional-looking thumbnails for you in a fraction of that time with no design skills needed. So, in this video,
[00:14] process, and then I'll run some of my old thumbnails through it to see how to make them better. The tool I'll be using to make everything is called Pixels, which is specifically built for YouTube thumbnails. So, it's not just another
[00:26] general image generator. If you want to follow along, I've left a link to it in with the part that surprised me the most, which is how little you actually have to do to get a finished thumbnail. Once you log in, just click on features
[00:39] at the top. To make a thumbnail from scratch, just click on prompt, which thumbnails. Then, you just type with simple words your core idea and generate. For the first one, I'll ask for a night raid of Bin Laden similar to
[00:52] those documentary fern videos rendered in a 3D Blender style with one figure in can find this exact prompt, along with the full workflow, down in the description completely for free, so you can follow along in real time. And
[01:06] need to overexplain everything in your prompt. Pixels understands thumbnail psychology, so just by explaining the idea, the style, and maybe some coloring, you'll get back something that works. We get back this image, which
[01:19] works well for a documentary video. It's clean, readable from a small screen, and the colors are eye-catching. You understand the whole situation before very important because a good thumbnail needs to get the message across in
[01:32] seconds before people scroll right past. Pixels gives an overall score to every thumbnail it makes, and this one comes back at 97 with a high rating on the clarity score, which means that it's instantly recognizable. For the next
[01:45] style, a how-to fix your posture thumbnail with the before in red and the after in green. And it splits the frame straight down the middle, the bad corrected one on the green. There's no text on it anywhere and you still know
[01:59] exactly what the video is about. That's the whole reason the before and after format keeps working. You read the two colors and you already know which one is the problem and which one is the fix. This one comes back at 95 and its
[02:11] virality is the highest of all the categories. So this has potential to finished thumbnails in just two sentences and I haven't opened a design tool once. And this method works for simpler concepts, but usually you need
[02:24] it something a lot harder and see if it still comes back with something you'd actually put on a video. This time I'll ask for a horror style kids drawing of a narcissist being exposed for what he is with the text [music] narcissist exposed
[02:38] on it. Notice I'm asking for text now, which is where most image generators that aren't properly rendered. We get back this pencil drawing of a man with half his face coming away and the text is spelled correctly and placed exactly
[02:52] where it should be. The style is spot-on for a psychology or a true story video, but this one only comes back at 76, which is way below anything we've made so far. So click on the score and you get the full breakdown, which is five
[03:04] different things: virality, clarity, idea, curiosity, and emotion. The lowest curiosity, and the emotion and they all come down to the same reason. It's a unsettling enough to attract attention, but it doesn't fully connect. There's no
[03:19] curiosity gap and both parts of the face look scary. A concept like this only works when the two halves of the image are showing you opposite things because that difference is what will make people click. So instead of rewriting the
[03:31] prompt, I'll click one click fix, which rebuilds it without you typing anything. It just gets the feedback from its own analysis. And now the mask is smiling while the face underneath it is falling apart. So it tells a complete story.
[03:43] That one change is what makes the whole thing work because you can see what he's in the same picture and you want to know which one is real. It also cropped in tighter, so those two faces fill the frame, which is what keeps it readable
[03:56] when it's tiny on a phone. This one comes back at 94 with a high clarity rating. It's the exact same idea, and it went from something I'd scroll past to Everything so far started from scratch,
[04:09] that's already proven to work and make it in your style? That's what recreate is for. And out of everything in here, it's the one I'd end up using the most. So, click on features again, and this time go to recreate. You paste in the
[04:22] link of a YouTube video whose thumbnail already works, or you upload an image straight from your computer, and then you type what you want changed. And this thinking about a thumbnail, because you're not copying someone's picture.
[04:35] proved itself and putting your own subject inside it. The format is what's doing the work, not the person in it. So, I'll paste in a video that already did well with this thumbnail, and my edit is simple. Change the man to Conor
[04:48] McGregor, and change that orange 659 text to a red 420. And this one comes back at 98, which is the highest we've had. The biggest thing here is that face, because a recognizable person will get clicks before anything else. Then
[05:01] the expression does the rest. That pain and disbelief, it's dramatic but not overacted, which is what most people get wrong when they pull a shocked face for a thumbnail. The score also doesn't only look at the image. It takes your title
[05:13] with it, because that's what people judge together in the feed. And with a famous next to an amount of money that small, so you want to know how he ended up there. That's the same curiosity gap we built before, which is something a
[05:26] good thumbnail always needs. So, I took a thumbnail that already worked, changed two things about it, and got the highest score of anything we've made. If you're works well. But what if you want to recreate an existing thumbnail, but this
[05:39] time use your own face on it? That's what a persona is for, and it's Pixel's unique feature for creating an avatar based on yourself. You upload a few on them, and from then on, you can select your persona on anything you
[05:53] characters, too. So, if you've created an AI influencer, you can upload a few pictures of them, and then you can use them in your thumbnails. So, I'll upload a few pictures of myself, name the persona Yuri, and click generate. And
[06:06] once it's done, I've got an avatar that looks just like me, ready to be placed in any project without me ever filming my real face. Now, I'll go back into recreate, paste a product review video whose thumbnail already worked, and this
[06:18] time my edit is a bit longer. Change the text that says 20-second instant toothbrush to scam, drop the price from 300 to 997, make the green rectangle red, and select my persona. I also told it not to change anything else which
[06:32] matters, because without that, these tools start redesigning the parts you thumbnail with me testing the product on one side and the product on its own with the verdict and the price on the other.
[06:44] what the answer is before you've read the title. But the smartest thing about it is that it gives you the verdict and hides the reason. You know it's a scam, you have no idea why it costs $997 or what it promised to do, and that
[06:57] missing part is what actually gets the click. Me actually testing it is important, too, because that accusation coming from someone holding the product is far more believable, and people want to see the reaction. All the changes I
[07:09] asked for are applied, and they deliver a completely new message in the thumbnail. This one comes back at 94 with a high virality rating, so most of that is the price shock and that one word. The lowest of the five is the
[07:21] reaction would hit harder, which like this, your face has to be as intense as the claim. The best part is that I look exactly like myself inside a thumbnail format that's already been
[07:34] proven to work. All these are brand new, and no matter the stats, we have no idea how they'll actually perform until we upload them. But Pixels allows you to analyze an existing thumbnail that you know didn't perform well, and it tells
[07:46] you exactly what to fix. So, let's do that with two thumbnails I made myself, performed. Click on features and go to Pixel Score. Then, upload your thumbnail and your title because it reads the two of them together the way a viewer does.
[07:59] whether your phone is listening to you, and it comes back at 44. So, let's actually go through what those five scores mean because this is the part that tells you what to fix. Virality is how far it can reach past the people who
[08:12] already watch you. Mine got a low rating because it never used Apple at all, and the mood is far too soft for a topic that's supposed to feel invasive. Then, someone understands it in one glance. A medium rating here because the text
[08:26] repeats my title word for word instead of adding anything new, and the phone in it is so small that nothing on there is the focal point. Idea is how bold the concept is, and a medium rating is what you get when you ask a question instead
[08:39] of stating something. Stating a truth feels a lot more engaging because the viewer will want to actually confirm it. Curiosity came back at low. We've thumbnail needs to create curiosity, and for this one, the cartoonish look
[08:52] that's supposed to be intimidating. Emotion rating is low, which is whether the feeling is believable. There's supposed to be fear on his face, and it click one-click fix [music] and hit fix my thumbnail, and it rebuilds the whole
[09:06] thing off its own analysis without me writing a single prompt. It comes back with Steve Jobs, a phone screen that says recording with a timer running and a red waveform behind him. It threw out the generic phone and used the one face
[09:18] everyone already connects to Apple, then made him the one doing the listening, which turns my question into an accusation. The running timer is what ignore, and something that looks like
[09:30] the emotion, instead of trying to portray fear, it's just a calm face next far more intimidating than a shocked expression would have been. This one comes back at 99, [music] up from 44, with the emotion going from
[09:45] a low rating to a high rating. The second one is worse. It's a video about thumbnail was a black background with a Reddit search bar typed into it. That one scores 33. The curiosity came back with a low rating, and the reason is
[09:59] anything on its own other than the fact that this is a Reddit video. It tells story, so there's nothing left to wonder about. But the interesting one is the clarity, with a high rating, the highest of its five. So, it was perfectly clear
[10:14] and readable, but it fails at everything else. Clear doesn't immediately mean two, the thumbnail will most likely fail. The emotion has a low rating, and feel anything about. So, I'll run this one through one click fix, as well. This
[10:29] one is a huge improvement. There's a masked figure in a dark room holding a syringe with the words dark side [music] of Reddit next to it. It stopped showing Reddit and started showing what that dark side could possibly be like. It's
[10:41] attention, creates curiosity, and causes stronger emotions. It's also clear that the only way to find out what's happening is to watch the video, which is why curiosity went from a low rating to a high rating. The emotion went from
[10:55] a low rating to a high rating, too, which is the highest emotion score from thumbnails, which were proven to not work, reached an overall score of almost 100, and the difference between them is clear, and it didn't require any
[11:08] designing skills from me. Pixels handled everything in seconds. With this tool, you'll never have to spend time struggling in Photoshop or Canva again. professional thumbnails in just a few minutes, use the link in the description
[11:21] to sign up to Pixels. Thanks for watching, and I'll see you in the next watching, and I'll see you in the next one.