How I Create Gorgeous Thumbnails with AI
Crafting the perfect thumbnail has become an art-form in itself and can significantly influence the success of your video content. Even YouTube heavyweights like Mr Beast claim the importance of thumbnails on overall views. However, getting the perfect photo with the perfect backdrop can be challenging for a baby-tuber with zero budget.
In strolls the AI revolution and suddenly, a plethora of gorgeous backgrounds and elements showing exactly what you need for your thumbnail, are a mere text prompt away.
In today’s blog I’ll walk you through how I’m harnessing the power of AI to create attention grabbing, beautiful thumbnails that integrate perfectly with my personal brand and help me stand out. Hopefully you’ll gain a few things that help you on your way to using AI and putting it to use.
(I mean, why not!? AI companies have already generated millions in revenue from the work of artists and musicians—without seeking permission, so it seems only fair to ‘turn the tables’ a bit and utilize AI to our advantage.)
Using AI: Step by Step Guide
Here is my breakdown of how I created the image above using AI.
Step One – Take a starting photo
I took the photo below while I was filming my music video and had the backdrop and lights set up with my hair and make up already done for the music video. I just put my camera on timer and posed a few times until I had something I liked. This was the result:
Step Two – Cut the photo out in Photoshop
The truth is, I’ve actually got some pretty advanced graphic design skills – all self taught mind you, so there’s no reason you can’t learn too if you are so inclined. However, Photoshop makes it SUPER EASY to cut out a picture these days. It has AI embedded into the software so it detects a person and selects them from the background with a single click.
Then you just invert the selection, delete the background and save the image as a PNG to preserve the transparency.
I think Canva and some phone apps also offer this technology if you don’t have photoshop. Just Google around till you find one. Here is the same photo but cut out:
Step Three – Make the background with AI
You have several options now when it comes to AI image generation tools. I used to use MidJourney but you have to pay for it and it’s cumbersome to use.
Along came the latest ChatGPT update and they now have Dall-E integrated right inside the Chat platform. (Wohoo!) Bear in mind, you have to be a paid user to access the Beta features (Beta just means ‘in testing still but available to the public so we can iron out the remaining bugs before it officially goes live.’) However, considering I was already paying for ChatGPT-4, this was a fabulous upgrade for no extra, plus it meant I could ditch MidJourney and save money there.
The trick in getting AI to generate anything decent is in your prompt. ‘Prompt Craft’ they call it. So the way I’ve learned to get the best out AI is to give it some context before going in for the ask.
So I talk to it like I’m explaining what I’m creating to a friend and say “Hi Chat, I’m creating a music video for the song What A Wonderful World. It’s got this romantic vibe and I need a thumbnail for YouTube to go with the video.” Funnily enough, Chat likes this and then when you ask it for the picture you get WAY better results than just diving in with a demand for ‘make something pretty and whimsical’.
You might also need to prompt it a couple of times to get it right and you can use previous images it creates as references. Ie “I like the first image. Can you create something like this but with more pink”.
NOTE: It still doesn’t create an exact replica just with more pink, but it does create something similar… with more pink.
Also – don’t forget to ask for the aspect ratio. (This is graphic designer speak for image dimensions). You can just say ‘use aspect ratio 16:9’ in Dall-E to get a landscape thumbnail size.
Here’s what I got:
Step Four – Add the butterfly using AI inside Photoshop
So now I have my cut-out photo of me, and a pretty background. I put the background layer beneath my photo, arrange my photo to the size and spot I like, then I merge the two images together.
Photoshop now has AI image generation built into it too…. BUT in my experience, I find that it’s good for small things like adding a butterfly, but not as good as creating the whole background (annoying). So that’s why I create the background using Dall-E first.
(I’m sure Photoshop will improve with time, then you can just do the whole thing there, which will be awesome.)
To create the butterfly, you literally just make a selection around the spot where the butterfly needs to be (in this case, sitting on my finger like I’m some Disney Princess that can summon pretty creatures to her in a heartbeat)… add in your prompt ‘Pretty butterfly’… and BOOM! It gives you three options.
Here’s a different butterfly option it gave me:
Step Five – Blend it all together in Adobe Lightroom
Ok, so this last step is where the real magic comes in (if you ask me). It’s the final icing on the cake that makes the image look cohesive. Like all the elements I’ve dumped in belong together.
Enter Adobe Lightroom.
I’m an avid Adobe user so I don’t know if Canva or apps can do this too. I’m sure there’s something out there that can.
So there are stages to this step:
A – Slap on your favourite preset
I have a few presets already saved in Lightroom that I like to use. So I just try them out until I find one that works the best and I just adjust the overall grain or exposure/ contrast from there slightly.
B – Create a mask to select your subject
Then increase the vibrancy or exposure just a teeny bit. For some reason, lightroom isn’t quite as good as Photoshop at selecting the subject so I use the brush tool to add bits to the mask that it’s left out (like my whole hand) ref below:
C – Duplicate the mask and invert it to select the background
I like to blur the background a bit and darken it so my subject (me) really POPS. Oh and I duplicate the mask instead of creating a new one for the background, because I’ve already done all the work of fixing the selection with the brush tool. If I started fresh, I’d have to do it all over again. (Yawn)
D – Add linear masks for lighting depth
This may or may not be the pro-photographer way of doing things, but I add two linear masks. One coming from the direction I’m facing – and I make this a bit lighter. Then another one behind me, which I make a bit darker.
This again, helps my photo POP and adds a bit of depth to the image and generally blend everything in seamlessly. Ref:
Then that’s pretty much it! Now I have a fabulous looking thumbnail to use for a YouTube video, or promotional picture etc.
I’ve even started making a few 1920’s jazz club backgrounds in AI that I can use later if they decide to increase the price and I become too cheap to pay for it! hahaha.
So I hope you found this blog post helpful. I know that if you’re new to graphic design, it can all look a bit daunting, but it really is a whole new world of fun creativity, so I’d encourage you to dive in and learn to use some of these tools if you feel like it.
Hit me up in the comments below if you need me to clarify anything and I’ll do my best to help out. Oh and below is my music vid for What A Wonderful World! (the music is all me, 100% real of course!!!) – I can’t believe we need to clarify that nowadays. How times have changed.