Why I stopped using AI-generated visuals?
When I started using AI for visuals, I became too excited about its capabilities. DALL-E was by far the most exciting one. At first, I used it for hours just for fun.
Then I started writing here on Medium. I used AI-generated visuals for my articles. I thought it was an amazing idea, creating visuals without having to design or take a picture for a specific topic compared to using photos from Unsplash.
Yet, after spending enough time on both DALL-E and Medium, I realized many people use AI-generated visuals on Medium, so they appear rather cheap-looking than creative. And I judged the effort put into the article just because of its visuals. At the same time, you cannot generally avoid AI putting something inhuman or illogical in the picture, whether it is an extra arm, an excessively large amount of irrelevant objects on the floor, and such, and you don’t have much control over the visual.
Here is an example when I tried to recreate a scene where Carrie Bradshaw was writing an article in her apartment, but I ended up with complete chaos while attempting to teach the AI better. Trying to improve the visual by providing more context generally doesn’t work. It is better if you can achieve success with a single command.
By understanding how it works and its flaws, I realised I actually skip reading the articles with AI-generated visuals. They literally started to annoy me!
Successful attempts with AI-generated visuals
Yet, I must give it credit. Sometimes, when I am lucky, it creates really good ones as well. Here are a few examples:
This one was created to depict an overwhelmed employee in an office environment. The woman is not unrealistically beautiful (though she generally is), and the rest of the photo is super realistic, too.
I love this one. I created it for my book review of Haruki Murakami’s Dance Dance Dance. This photo was supposed to be unrealistic. AI literally generated “the one”. And for this one, my commands actually worked, and AI was able to deliver what I asked for.
This one is a Turkish dish: “Kıymalı Patates.” The only thing wrong with this photo was yogurt with paprika on the side. We don’t eat such a thing. And actually, the onions look kind of large and raw, don’t they? Yet, I loved its traditional look.
This one is created for an article about Victor Horta, who is a pioneer architect of the Art Nouveau movement. This is his house, which serves as a museum in Brussels today.
And then I gave up. It literally took too much of my time to fix what’s wrong with the visuals just by commanding. I found it revolutionary yet cumbersome. I think, in a few years, we will be more flexible in commanding the picture. Yet, then it will raise the problem regarding the visual’s authenticity, which is actually kind of scary!
