27 Comments
User's avatar
SipAndRamble's avatar

This is wonderful but what if you don't have original ones to go off from?

Expand full comment
Luan Doan's avatar

That’s a great question, here are a few things you can try:

- If you’re starting from scratch and want solid branding, you could hire a designer to help create your visual identity

- You can also look at your current thumbnails, pick the one you like most, and build a consistent style from there

- Or try working with ChatGPT to help you design a visual branding guide - once you’ve got that, you can use it to keep your thumbnails consistent

Expand full comment
SipAndRamble's avatar

Very good suggestions. I will try the last one. Should be fun. Thank you.

Expand full comment
Larry C. Brown's avatar

I’ve been thinking about creating a consistent thumbnail style for a while. This is exactly the information I needed to do that. Thank you!

Expand full comment
Luan Doan's avatar

That's great, Larry. Very glad it resonates with you.

Expand full comment
Karo (Product with Attitude)'s avatar

Great post! My ugly doodles and I are very proud to be a part of this experiment🤗

Expand full comment
Luan Doan's avatar

I find it very cute 🤗. Thanks again, Karo, for joining this.

Expand full comment
William Meller's avatar

That was really nice to learn!

Expand full comment
Luan Doan's avatar

Thanks, William.

Expand full comment
Amal Shah's avatar

Surprised I haven't thought of this consistently. I've done it a few times but not as a Custom GPT.

Thanks for this.

Expand full comment
Luan Doan's avatar

Yes, Amal. The idea is to create a reusable system focused on repetitive work, this way, we can use AI more effectively and save more time.

Expand full comment
Doug Keefe | Build Fast AI's avatar

Love it!!

Expand full comment
Luan Doan's avatar

Thanks, Doug!

Expand full comment
Kamil Banc's avatar

Good concepts, i’m surprised how much ChatGPT has gained when it comes to market share in photo and image generation I used to use 2 to 4 different tools and now I just use one

Expand full comment
Luan Doan's avatar

Yes, Kamil. With the recent GPT-4o image generation model, OpenAI now provides the ability to create accurate images with text based on prompts. While there are still some limitations, I think we’re just getting started and there are many interesting things to come. Do you also create thumbnails using ChatGPT?

Btw, I’ve been following your newsletter for a while, and I’m really impressed with the visual images you include in your posts, consistently styled and clearly illustrating the content. Do you create them manually using design software?

Expand full comment
Kamil Banc's avatar

Thanks Luan, 90% of my thumbnails are made with 4o. I have a project with a style guide set up for that and a work flow that’s working.

Infographics are mostly done with napkin.ai

Expand full comment
Luan Doan's avatar

That's great, Kamil. Thanks for the information about napkin ai, I'll give it a try.

Expand full comment
Adi Sharma's avatar

I was using this without realizing its importance. Thanks for the research and laying it down so concretely, Luan!

Expand full comment
Luan Doan's avatar

Thanks, Adi. Glad it helpful for you.

Expand full comment
TechTiff's avatar

Awesome article, Luan! You seriously nailed the reverse-engineering approach!

Expand full comment
Luan Doan's avatar

Thanks, Tiff

Expand full comment
Daria Cupareanu's avatar

This was really cool, Luan!! I think this kind of reverse-engineering is such a foundational practice for generating consistent, on-brand content. Obviously a few more tweaks are still needed in some cases, but the approach itself is super powerful. Reverse engineering is such a great hack that can be applied to so many use cases.

Expand full comment
Luan Doan's avatar

Absolutely, Daria! What’s exciting is how AI is making this kind of reverse engineering way more accessible, even for solo creators. We can now quickly spot patterns, dissect structure, and learn a lot from them.

Expand full comment
Jenny Ouyang's avatar

Love the concept of reverse-engineering those thumbnail styles, some of them definitely hit the mark!

Expand full comment
Luan Doan's avatar

Yes, it still needs improvement, this is just a starting idea. If we take more time to test and optimize, we could actually create a usable system for generating images for articles and social media.

Expand full comment
Wyndo's avatar

Love this framework!

And yes the secret is to reverse-engineer your thumbnail by dissolving each key elements of your visual into a prompt that can be replicated across multiple future posts.

I think I like our combination visual, looks like star wars haha

Expand full comment
Luan Doan's avatar

Haha, yes, I like the yellow stripe detail. Very magical.

Expand full comment