Years ago, I was very enthusiastic about generating kaleidoscopes from my images, publishing at least one series per week, with over 100 posts.
@seckorama/kaleidoscopes-102-expectations
Then my enthusiasm waned and, like everyone else, I started experimenting with AI and devoted myself more to other things.
But, these days, I have revived my interest in kaleidoscopes. I still like these geometric color structures. My guiding principle in this has always been the fact that a kaleidoscope is just a different perspective on reality and nothing extraterrestrial.
Here are four of my photos of flowers, on which I used the Paint.Net DPL distortion kaleidoscope effect and changed a couple of parameters. This allowed me to create different geometric and color structures from a single photo. I used to use the Photoshop Mehdi kaleidoscope filter, but after switching from 32-bit Windows 10 to 64-bit Windows 11, it no longer works for me.
These images were created using filters and effects in graphics programs; I did not utilize AI.
I later used AI because I wanted to see how much AI had advanced in this area. I then added all four photos to Night Cafe Creator and tried using the kaleidoscope effect with different AI models there as well. But more on that later.
Cactus
And this is a photo of a cactus from which I generate kaleidoscopes.
Finally, here is the AI result on the prompt: kaleidoscope, 6x mirroring, prism.
Model: HiDream I1 Dev
Roses
Original photo
AI-generated image from original and text prompt: kaleidoscope, 6x mirroring, prism.
Model: Flux
Giant bellflower
Original photo
AI-generated image on the same text prompt as above.
Model: Dreamshaper XL Lightning
The Sweet William
Original photo
AI-generated image on the same text prompt as above.
Model: Ideogram 2a Turbo
So, what are the conclusions based on what we have seen?
I will start, and then you can comment.
AI is more or less hallucinating and has no idea what a kaleidoscope is :) 😂
I used four different models to generate images, and in none of them did I get even a slightly adequate result compared to what plug-ins and add-ons for graphics programs generate, far from it.
It's such a narrow niche that they didn't learn anything from the photos.
And I didn't just do this in Night Cafe Creator.
Before that, I also tried Grok, Gemini, and Qwen, and it was even worse. All three generated something that had nothing to do with the original photo or kaleidoscopes.
So I'll stick with graphics programs; AI is useless for kaleidoscopes. I may be wrong.
Thank you for your attention.
And here's something else:
👉 @commentrewarder waiting for your comments.
👉 Support and vote for those exceptional witnesses:
@good-karma, @liotes, @fbslo and @detlev.witness.
They appreciate any help you can provide.
NFT Showroom
3Speak Channel
Let's Make A Collage
Beer Tasting
New Tunes
Foodies
