Why Making Images with AI Burned Me Out

AI Can Generate Images — Just Not the Ones I Want

AI image generation is supposed to save time and energy.
At least, that’s what I believed.
But after waiting minutes for results that felt almost right—and repeating that process over and over—I realized something was wrong.
I wasn’t getting better images.
I was just getting tired.

I was using AI to create top images for this blog.

At first, I wanted simple, light, and slightly comical illustrations, so I gave the AI instructions accordingly—but it turned out to be much harder than I expected.
No matter how many times I tried, everything felt off. Like, “This is not it at all.”

“Seeing Is Believing” Didn’t Work Either

They say “seeing is believing,” so I even showed the AI examples of illustrations I liked and said, “Make it this kind of style.”
Still, it didn’t work.

Maybe because of copyright restrictions, it can’t really use real illustrations as references—every result came out with a completely different nuance.

Maybe drawing them myself would be faster. But it wouldn’t. It would take even longer—and I can’t even draw.

“Same Style” Does NOT Mean “Same Illustration”

But then, once, I finally got an illustration that was pretty close to what I wanted.
So I showed that image to the AI and said, “Please use this same kind of touch.”

And what did it do?

It created an illustration with almost the exact same composition as the reference.
I said “the same style,” not “make the exact same illustration.” 💢

After that, the AI never once managed to recreate the same ‘style.’

Waiting Minutes Just to Get Something Weird

What really frustrated me is that AI image generation isn’t instant.
You wait one or two minutes, only to get a weird image—and if you repeat that process over and over, it becomes exhausting very quickly.

Even when I gave extremely detailed instructions, the AI would make totally irrelevant changes—or sometimes change nothing at all.
That kind of thing happened constantly, and eventually, I just got sick of making illustrations altogether.

What I Actually Wanted

For reference, this is the kind of illustration I originally had in mind.

Doodle-style illustration showing an angry customer shouting “Customer = God” at a tired shop worker, representing issues in Japanese service culture

“Customer = God”? How a Strange Phrase Damaged Japanese Service Culture

Something simple, a little doodle-like, and slightly comical.

My blog sometimes deals with heavy topics, such as social issues in Japan, so I wanted the illustrations to feel light and not too serious overall.

But the AI just wouldn’t work the way I imagined.

For example, when I wrote an article about Akabane, Tokyo—a neighborhood known for its many izakaya(a casual Japanese pub serving food and drinks)—I wanted an illustration of a drunk person.
The first result wasn’t what I had in mind, so I asked for something more comical and goofy.
And then the AI gave me this old man illustration.

AI-generated cartoon illustration of a drunk man slumped over a table, intended to represent a comical izakaya scene in Akabane, Tokyo

Why is he sticking his tongue out and panting like that…?
What is he even excited about?
He just looks like a creep.

Who on earth would want to visit Akabane after seeing this?!!

Giving Up on AI Illustrations

At that point, my frustration hit its peak, and I gave up on AI-generated illustrations altogether.

So I switched to generating photos instead.

With photos, I often get something decent on the first try.
That’s why I decided to use photos as top images.
Though, to be fair, if you look closely, the photos still have plenty of weird details.

For example, this image was generated for an article themed around
“The Wildest Customers I’ve Ever Met in Japan”
When I asked the AI to “create a photo of a strange customer,” this is what it gave me.

AI-generated photo meant to depict a strange customer, but resembling a suspicious man in a dark hallway

That’s not a strange customer—
that’s a criminal!!!!
Does anyone actually look at an image like this and think, ‘I want to read this article’?

Do We Really Need a Top Image?

Still, depending on the topic, even photos can be difficult.
Sometimes I don’t even know what kind of situation to depict, and for manga-related articles, I can’t use images at all because of copyright issues.

And then it hit me—do we really need top images in the first place?
Isn’t it totally fine not to have one?
There are plenty of popular blogs out there with no images at all.

So I decided that when nothing comes to mind, I just won’t use an image.

Honestly, I probably should have done that from the start.

Similar Posts

Leave a Reply