Artificial Intelligence is currently one of the most discussed and controversial topics in the creative industry. Especially for illustrators, AI raises many questions, fears, and also curiosity. Some see it as a threat, others as a useful tool. Instead of taking a clear position right away, I decided to test AI illustration tools myself and see what they can (and cannot) do. The goal of this article is to explore it through a practical experiment and see what models are the best for illustrations.
Between Fear and Opportunity
I have already written a blogpost on this topic. However, for the purpose of completeness, I would like to reiterate the most important points here. AI in illustration comes with many concerns, and these concerns are valid.
On the negative side, many AI tools are trained on artworks without the consent of the original artists, which raises ethical concerns. Many illustrators fear job loss or that their work will lose value. AI also tends to copy existing styles without truly understanding their meaning.
On the positive side, AI can be useful when used carefully. It allows fast idea generation, quick experiments, and can support inspiration and composition during the creative process.
This mix of risks and possibilities is why AI remains such a challenging topic in illustration.
The Experiment: Recreating My Illustration with AI
For this experiment, I used one of my own illustrations, which I created manually in Procreate. This illustration served as the reference and starting point. The idea was to see:
- how close the AI results come to my original illustration
- how well the style, colors, and mood are translated
- where AI performs well and where it clearly fails
From my experiences using AI Genereators, its easier to let AI write the Prombt for the gererated picture instead of write it by myself. So I gave Chat GPT my Illustration and asked him to describe this illustration in form, color an style for an AI Generator Prombt.
This is how it turned out:
A surreal hand-drawn illustration of a ceramic teapot filled with water like a small aquarium. Inside the teapot, tiny fish and aquatic plants are floating gently. From the spout of the teapot, a soft pink stream flows into the air, and two larger pink fish are swimming within this flowing stream. Dark green textured background, soft pastel color palette with turquoise, pink, and muted grey tones. Whimsical, poetic atmosphere, clean linework, subtle grain texture, storybook illustration style, calm and dreamy mood.
Let’s start the AI Competition
For this experiment, I tested several popular AI image generator models mainly from Firefly. Here are the results and my opinion on it:
- Firefly Image 4 Ultra: it dosen’t make sense at all, the style is quite mixed up (3d and 2d elements), don’t understand the stain in the backgorund, at least the teapot looks quite nice with the aquarium in it
- Chat GPT Iamge 1.5: I think thats one of the best results when it comes to similarity, the style is illustrative and handmade, it actually make sense, but I don’t like the layout it seems kind of lost and random, the fishis are nice
- Gemini 3 with Nano Banana Pro: This modell is usually one of my favorites, I like the artsy style, I think its the best result in gerneral which you could actually use for something (or work further on that), It’s all kind of random and not exactly like the prombt but the style fits pretty well
- Flux 1.1 Ultra Raw: The only thing I like here is the bold and graphic style, without the fish in the air and the tiny smoke-thingy it could look nice
- Runway Gen-4 Image: What the hell is happing here?! But to be honest if there were no weird eel-ghost there, I would choose that as the winner picture. I really like the style, it looks very editoral-ish and artsy, even though its the most diffrent compared to the original one
- Leonardo AI: I tried this website for the first time as I read it should work well for illustrations, turns out: not really. I think its together with the Firefly Ultra its my least favorite, the style is kind of nice (maybe for posters) but it dosen’t make sense
I thought this experimetn was really fun, but I am not completley happy with it at this point. So I decided to take my favorite model and work on the gererated picture until I am happy with the result. Since I really liked the Style of the Runway Gen-4 Image Generator I chose to go with that one, even though the compostions is hilarious. But I see the most potentiol here to get a nice and artsy result.
But after half an hour promting by myself and chat gpt, the picture below was the best result I could get. So I gave Runway up again and switch to my usual favorite Nano Banana again.
I tried hard to create a nice result, but the one from Nano Banana was the best I could get. Its kinda nice, but the artsy and handdrawn style is totally missing. It always seems easier than it is to prombt something nice, but actually its pretty hard and time consuming. Probably I could get a better result than this, but I dont wanted to waste so much time (and of course energy) to create those images. And I think thats the whole point: It’s kind of fun to prombt something until it doesn’t work out the way we want or it dosen’t match our expectations. It was interesting and fun to create these images and see how differently each generator works, but it was not as enjoyable as painting the illustration myself. In the end all the created images are looking like AI images. I can’t say exactly why – maybe the too clean lines, the over perfectionsim or the plain textures – but you can see it in all of them. They have no soul and no personality. I don’t want to be an AI-hater, because I am sure that we all have to use it in the future, but with real art at least for me it can’t compete and won’t ever.
I am sure that for some tasks or areas, it makes sense, especially if it doesn’t matter much which illustrations are used. It can save a lot of time and produce beautiful results. Perhaps there are also ways to combine analog illustration and AI, which would certainly be interesting to test. But all in all, I believe and hope that human illustration and art will always remain an important and irreplaceable part of our lives.
Because Art without soul is like a book without words.
