Funny AI Rendering Fails
Jan. 14th, 2026 05:41 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
canyonwalker2025 was the Year of AI. Content generated by AI started popping up everywhere. I even used it a bit myself. But just a bit, because one thing that was blindingly obvious in the Year of AI— obvious to anyone really paying attention, anyway— is that AI can produce some laughably silly results.
For example, some of my colleagues went big into using AI image generation to illustrate slide presentations a few months ago. I couldn't help but point out, publicly, when a person (in the picture) who was the subject of the story had, say, 3 arms, or when they had a laptop showing our product on its screen while the screen was bent around the wrong way. Maybe that makes me a bad person. But I've always been the one dumbass who, when the emperor strides onto the stage stark-fucking-naked, nudges people next to me and says out loud, "Look, the emperor's wearing no clothes!" And more to the point here, if we don't object to AI slop right now, its' going to become normalized and we're going to be completely inundated with it in 2026.
Anyway, this journal entry isn't supposed to be a screed against AI. I'm writing to share some in-the-know humor about some of the funny results AI image gen gives us.
A few months ago I used Google's Gemini image gen to illustrate panels for a story I wrote on my blog. It's the one I only finally finished yesterday: The Mystery of the Church Up the Hill. One of the images I created was of my father painting the inside of the church he attended decades ago. I wrote a prompt like
And the first result was....
Disco Jesus! 🤣
I literally gave this prompt next:
To its credit, the image generator came back with a new image that did not have the son of god painting his own church after rising from the dead inside a vintage clothing shop. 🤣
Ultimately there were more things wrong with the pic than just "My dad doesn't look like Jesus", so I prompted the AI to start over. On my second try I used a few more terms to describe the aspects of the scene I thought were most salient. I got the image I used in the story I shared yesterday.
Was the final image I went with at all like the church my dad painted? No. But it conveyed the parts of the story I thought were appropriate. Including a few key elements of my dad's appearance: age, body shape, hair color, and glasses. One thing I couldn't get right in a handful of prompts was managing to dress my dad like a dork from the early 70s. Gemini kept taking the "early 1970s" prompt as making my dad look like a dork who dressed up to go disco dancing. Though I can see now that Dad would've looked pretty sharp— for a dork— in Chelsea boots and a leather vest!
For example, some of my colleagues went big into using AI image generation to illustrate slide presentations a few months ago. I couldn't help but point out, publicly, when a person (in the picture) who was the subject of the story had, say, 3 arms, or when they had a laptop showing our product on its screen while the screen was bent around the wrong way. Maybe that makes me a bad person. But I've always been the one dumbass who, when the emperor strides onto the stage stark-fucking-naked, nudges people next to me and says out loud, "Look, the emperor's wearing no clothes!" And more to the point here, if we don't object to AI slop right now, its' going to become normalized and we're going to be completely inundated with it in 2026.
Anyway, this journal entry isn't supposed to be a screed against AI. I'm writing to share some in-the-know humor about some of the funny results AI image gen gives us.
A few months ago I used Google's Gemini image gen to illustrate panels for a story I wrote on my blog. It's the one I only finally finished yesterday: The Mystery of the Church Up the Hill. One of the images I created was of my father painting the inside of the church he attended decades ago. I wrote a prompt like
A man is painting the walls in a Catholic church. He is in his 30s and is dressed in clothes fashionable in the early 1970s.
And the first result was....
Disco Jesus! 🤣
I literally gave this prompt next:
The man is not Jesus.
To its credit, the image generator came back with a new image that did not have the son of god painting his own church after rising from the dead inside a vintage clothing shop. 🤣
Ultimately there were more things wrong with the pic than just "My dad doesn't look like Jesus", so I prompted the AI to start over. On my second try I used a few more terms to describe the aspects of the scene I thought were most salient. I got the image I used in the story I shared yesterday.
Was the final image I went with at all like the church my dad painted? No. But it conveyed the parts of the story I thought were appropriate. Including a few key elements of my dad's appearance: age, body shape, hair color, and glasses. One thing I couldn't get right in a handful of prompts was managing to dress my dad like a dork from the early 70s. Gemini kept taking the "early 1970s" prompt as making my dad look like a dork who dressed up to go disco dancing. Though I can see now that Dad would've looked pretty sharp— for a dork— in Chelsea boots and a leather vest!
