I thought AI would fast track my art. Instead, it was a slow detour back to the basics and continuous creative existential crises.
The Honey Moon Phase
At first, it was fun. I uploaded my hand-drawn line art, and the AI coloured it for me. Sometimes it nailed the palette. Sometimes it hallucinated, adding random people, buildings, or objects that didn’t exist. Sometimes it crashed my computer. Sometimes it crashed me. And sometimes, in peak absurdity, I used one AI tool to write a prompt for another AI tool. AI-to-AI. The closest artificial intelligence will get to intimacy.
I used the software provided by Artificial Studio: "Artificial Studio is a web-based AI platform that brings together over 40 pre-trained models for creating and editing multimedia content—all from one user-friendly interface. It lets you produce and manipulate images, videos, audio, and text using tools like image upscaling, video face-swaps, background removal, style transfer, text-to-video conversion, logo and packaging design, 3D image generation blah blah blah"
The Experiments
The piece I was working on was deeply personal, a hand-drawn image of long-gone businesses in Limerick, Ireland. I wanted nostalgia, authenticity, a certain handmade soul. AI had other ideas.
But it never quite got it right. It looked false, it took away my wobbles, my imperfections, my inaccuracies, my mistakes, it took away me. I was no longer the artist and the work didn't belong to me.
Below i'll show you my journey with Ai, from start to finish, I mainly used Open Ai Image 1, Bagel and Flux Kontext to colour the images with varying success and often used Chat Gpt4 to generate the prompts.
This was the image I had created, by hand with a pencil and fine liner and uploaded to Artificial Studio.
Prompt 1:
Comical. The text warped, the composition shifted. If this was the future of world domination, we’re safe for a while. AI really struggles with text. So don't reckon it will be marking students work anytime soon. I'm a teacher and struggle to decipher my students handwriting at the best of times.
Prompt 2:
Better. Cleaner buildings, good colour scheme. But it was too good. Slick in a way that didn’t feel like my style and had completely created an entirely new image.
Prompt 3:
Incremental progress. AI began to “understand” me, though it still removed details, added others, and forgot buildings. But it started to get interesting, it really was better than me. Numerous creative existential crisis's began than occur.
Prompt 4:
Again it was beginning to get better, but there was still something not right. It changed buildings, it left buildings out, and started to add details that I did not have in the original image. But it was good. I thought initially that I would just upload the image and within a minute I would have my outcome, by this stage I had probably spent about 15 hours of work just trying to craft the correct prompt to create the image. I could have spent those 15 hours just using my usual process, I began to wonder was this a worthwhile endeavour or just a rabbit hole of creative doubt.
The Final Prompt:
Then it came out with this. Gorgeous colours, nostalgic, warm, fuzzy, like a full spectrum hug from the past, this level of colour would have taken me weeks (the drawings had already taken a year). But AI kept adding people, cars, and bicycles I never asked for. AI loves to hallucinate, even when you tell it not too, possibly a contact high from all those Silicon Valley CEO's spending too much time sipping ayahuasca...
The Creative Guilt
That’s when the creative ethical dilemma hit. The AI was undeniably better than me at colour, something I’ve always preferred to avoid in favour of line, plain backgrounds and silhouettes. But if I released it, was I the creator or just a curator of an algorithm’s output?
I asked Chat GPT (yes, I’m aware of the irony) what to do. It suggested adding paper textures, fixing the text in Photoshop, and even recommended tools for increasing resolution making the image feel less AI generated. AI has AI’s back.
After countless iterations, 100 variations in total, I found myself asking the big question: Should I just fix this in Photoshop and release it as my own?
But I couldn’t claim it. Even after hours of work, it didn’t feel like mine and it wasn't, it was a combination of other artists work that the AI had scraped from the internet to produce something it thought I wanted.
Above is a print of an earlier work I had created, this was the first in the series of the 061 of Old. Done with pen and ink (and a touch of Photoshop).
What I Learned
Artificial Studio isn’t bad, and I can see in the future how it will be extremely useful at doing mundane tasks that would usually take you a few hours in Photoshop, but it’s not ready to justify a paid subscription. It often guesses what you want instead of delivering exactly what you ask for, and the interface changes frequently, like way too frequently, it's all over the place. More importantly, using it slowed down my creative process instead of speeding it up, I could have released the final print weeks earlier if I hadn’t gone down the AI rabbit hole.
That said, I’m glad I tried it. As an educator, I now have first-hand experience of AI’s strengths and weaknesses in the creative process. But right now, too many companies are rushing to monetise unfinished tech, selling hype rather than reliability.
100 AI prompts later and I decided to go with the image below. This is the second in the series of The 061 of Old and it felt right to keep that consistency between the prints. Like Photoshop in its early days, AI will likely become another tool in the artist’s kit, I just don't think i'm ready for it just yet, or maybe it's not ready for me.
Click on the print below and you can even zoom in to see all my mistakes! If you made it this far, thank you so much for reading.
Go raibh milé.
