GPT‑4o Image Generation in Action: Vintage Cookbook, Learning Book Page, and Comic Design
How to generate images with Chat GPT? Can AI really design a full illustrated cookbook, an educational page, or even a comic book in just minutes? 💡 With GPT-4o — absolutely.
I am writing this article on 4 April 2025. This is especially important to note in a world where technology is changing so quickly! I'm sure that one day I will write something with the exact time down to minutes (or seconds!), because a little later my text will seem really old.
It's been such an exciting time for everyone who's interested in AI and those who use it every day - it's been a rollercoaster of a week! 🚀
🤖 All of my social media feeds have been overwhelmed with images created by Open AI's new 4o image generator. ✨ A viral trend with Ghibli portraits. It's simple, you need the new GPT-4o, your photo and a best Prompt for ChatGPT 👇
Transform the uploaded image into a Studio Ghibli-style animated portrait. Keep the person's main features recognizable, but adapt them to Ghibli aesthetics: large expressive eyes, soft and rounded facial features, smooth skin, and a gentle, kind expression. Use a soft pastel color palette and add a whimsical, nature-inspired background with floating petals, soft lighting, and a dreamy atmosphere. Style it like a magical scene from My Neighbor Totoro or Spirited Away.
We're all so familiar with that struggle, right? When creatives are keen to come up with something visual and include some words in it, it can feel like a real uphill battle. It used to be almost impossible, and I feel for anyone who's struggled with it. You often had to edit text that was written poorly and full of mistakes. Thank goddess ChatGPT is a linguistic model!
I was really impressed by the ability to provide text in the prompt and get it in the picture without changes and in the right context – that's what I liked best about the new GPT-4o. I just had to get it to try it out myself, of course! Now, I'll share my steps with you in detail.
- Experiment 1: I created a page from a vintage-style recipe book
- Experiment 2: I created an educational book page for a biology lesson
- Experiment 3: I created a short comic strip about aliens
How to Generate Vintage Cookbook Illustrations with GPT-4o
To be honest, I love both new technologies and vintage. 🕰️ Living in Central Portugal, it's easy to love vintage - there are so many houses, cafes, crockery, old cars, and sometimes it seems like time has stopped and it's not later than the dark days of the Salazar dictatorship.
But actually, this is a new post-information era, and it's great! But sometimes you just want to dive into the vibe of the old days, right? 🍽️ That's why one of my projects, which I've come up with myself, is a collection of old recipes with pictures.
Another thought - sometimes, when I don't know what to cook anymore, and I'm not in the mood to go to a local restaurant, I ask for time to come up with a recipe for me. Or find one online. But to be honest, they're a bit monotonous: chicken, avocado, lettuce, healthy eating, and so on. But the smell of BAKED COTTAGE ROLL, made according to my great-grandmother's recipe, is exactly what I need.
📜But let's get to the point! How to generate images for my vintage cook book with Chat GPT? So today I asked GPT-4o to create one for me 👇
Please, create a vintage-style illustrated recipe book page featuring the recipe "Baked Cottage Roll". The design includes elegant, old-fashioned typography, delicate hand-drawn illustrations of ingredients (cottage roll, cloves, bay leaves, onion, brown sugar, cinnamon, egg yolk, orange juice, bread crumbs), and a parchment-style textured background. The text of the recipe is clearly readable in a retro serif font. The layout is charming and nostalgic, reminiscent of 1950s cookbooks. Include decorative borders and subtle watercolor elements. Warm, earthy color palette.
Recipe taken from Cook Book by First Lutheran Church (Winnipeg, Man.).
BAKED COTTAGE ROLL
2 lbs. cottage roll
4 whole cloves
2 bay leaves
8 cups water
1 onion sliced
Scrub the cottage roll well; put into a deep kettle; add the other ingredients and the water; bring slowly to a boil; simmer covered for two hours. Let cool in the liquid, remove to a roaster; cover with the following topping:
½ cup brown sugar
¼ cup dried bread crumbs
1 teaspoon cinnamon
1 egg yolk beaten
¼ teaspoon cloves
2 tablespoons orange juice
Mix the ingredients and spread over the top. Bake for 30 minutes, basting frequently.
— Mrs. J. Forsyth, Ericksdale, Man.
I was really pleased with the end result. The colours, drawings, style, and the way the text is written are all amazing. 👇
But it's not perfect. The word 'topping' appeared twice, so I tried to fix it, but it only got worse. This is just an example, and I don't want to make this vintage page perfect right now, so I decided not to waste time on further corrections.
You can read my short dialogue with the Chat GPT Image Generator at this link.
Chat GPT Image Generator for an Educational Book Page
When I wrote this post, I was planning to make an educational poster. 📚 But something went wrong, the stars aligned differently, and I made a mistake in my Chat GPT Image Generator request. I guess I got a bit carried away thinking about school biology classes and looking at plant cells under microscopes.
So I asked to generate a page from the school textbook. 💬 I didn't really explain myself very well at first, so the Chat GPT didn't understand what I was talking about. Then I decided that the info from Wikipedia was just right. 🧪 Things went better after that – GPT‑4o generated an illustration of a plant cell that looked more like something you'd see in a book.
Please create for me a realistic and educational textbook-style illustration of a plant cell, labeled with its key parts: cell wall, cell membrane, chloroplasts, cytoplasm, nucleus, vacuole, mitochondria, and ribosomes. The image should resemble a page from a school biology textbook for middle school students. The layout should include a clear, colorful diagram in the center or on the right side, with neat English labels and soft background. Style: clean, scientific, easy to read, with a slight 2D flat aesthetic suitable for printed educational material.
Then I decided to ask another question, which was more specific, but still didn't have a text component. The result was the same: the chat either generates a picture or text. So I generated a text block separately, making it deliberately large. I was curious to see what GPT‑4o would do with it. It took the easy way out and just cut off the excess. Here's the final result.
You can read this dialogue with the Chat GPT Image Generator at this link.
GPT‑4o: Create Comics with Just One Prompt!
What do you think of this idea? 👾 Just to create a comic with just one prompt! I gave it a go, and I'm pretty happy with how it turned out.
So, here's how to create a comic with GPT-4o.
First, it's obvious that the comic will be short. After all, the GPT-4o only generates one image per request, so you can't use it to tell a long story. But it's still possible (and important) to highlight parts of the comic that are more complete.
So, I've come up with a script and dialogues for a short comic strip with four chapters. 👽 My comic is about aliens, it has a few characters and short dialogues. Despite this simplicity, the chat had some problems.
We already know that to successfully generate images that include text, the text has to be written in the prompt.
🛸So, to generate my comic, I used such a prompt:
Create a 4-panel comic in a modern, colorful comic book style with clean line art and expressive characters. The theme is humorous sci-fi. Each panel includes dialogue in speech bubbles.
Title: "First Contact"
Panel 1: A flying saucer hovers over a dark, grassy field at night. Two cute but slightly grumpy aliens with big eyes and funky suits are looking down from the cockpit window.
Alien 1 (excited): “Finally! We’ve found intelligent life!”
Alien 2 (smirking): “Yep, the signal came from this planet!”
Panel 2: On the ground, a guy in a bathrobe is holding his phone up, taking a selfie with the spaceship in the background. He’s throwing up a peace sign and grinning.
Human: “Yo! Alien content for TikTok! #Aliens #NoFilter”
Panel 3: Inside the ship again. The aliens look at each other, clearly confused and slightly concerned.
Alien 1: “Did we jump the gun on that ‘intelligent’ part?”
Alien 2: “Maybe he’s their influencer-in-chief?”
Panel 4: The spaceship is flying away into the starry sky. The aliens are sitting back, looking tired.
Alien 1: “Let’s try Mars next. At least it’s quiet there...”
Alien 2: “And no dancing before invasions.”
Style reference: modern Western comics, similar to Invincible, Saga, or Scott Pilgrim – vibrant, slightly cartoony but detailed, with expressive faces and cinematic composition.
The first result was great! But GPT-4o got a bit confused with the human's hands and just drew a nonsense. I tried to point out the mistake, and the second option was also good, but not exactly according to my script, which was in the prompt. 🪐 In the end, I just reminded it to stick to the script, and the final result was great.
You can read our dialogue here.
So, if you've ever felt a bit stuck trying to combine visuals and text, now's your chance to explore. AI isn't perfect yet, but it's getting there, and it's more fun than ever to push the boundaries.
🎁 Don’t miss out on future prompt magic! Use code PROMPTMAGIC for 25% off your annual plan! Try out a prompt, tweak it, make it your own — and don't forget to enjoy the process. 🚀