A few days ago, I was chatting with a friend on Hive, and he was reminiscing about his early days on Hive. He started with flowers :)
Well, he wasn't the only one. I also started by posting photos of flowers. Even now, nine years later, I still do that occasionally. Flowers and plants, in fact, all of nature, are a beautiful subject for beginners and photography enthusiasts.
But then I became extremely obsessed with it: namely, from photos of flowers and, later, everything else, I began generating kaleidoscopes using graphics programs.
I was attracted by the colorfulness and psychedelic nature of the geometric patterns in kaleidoscopes. And what you see in them is always a matter of perspective.
During the debate, we also discussed the importance of engagement. Participating in communities and discussions with others, helping less experienced users, and so on.
When it came to kaleidoscopes, I took a different approach. I started searching Hive for photos to use to create interesting kaleidoscopes. If I found something suitable in a post, I contacted the authors and asked if I could generate a kaleidoscope from their images.
I don't remember anyone refusing to give me permission.
And now I am returning to my roots.
@ph1102/well-known-road-wednesday-walk
But I didn't stop there :)
To benefit the community with my experience, I went even further.
Half a year ago, I checked how AI specialized in image generation understands kaleidoscopes. The conclusion was that they hallucinate too much :)
@seckorama/kaleidoscopes-in-graphic-programs-and
Has there been any progress? I think so, but AI is still hallucinating. Unlike last time, I used general LLMs with the same prompt: "generate kaleidoscope."
Now, kaleidoscopes.
馃摲 馃槑 馃摲
Original photo
generated with Photoshop Mehdi Kaleidoscope Filter
generated with Paint.Net Distortion DPL Kaleidoscope Effect
generated with Grok (X)
generated with Qwen (Ali Baba)
generated with Gemini (Google)
Qwen was hallucinating too much :)
馃摲 馃槑 馃摲
Original photo
generated with Photoshop Mehdi Kaleidoscope Filter
generated with Paint.Net Distortion DPL Kaleidoscope Effect
generated with Grok (X)
generated with Qwen (Ali Baba)
generated with Gemini (Google)
馃摲 馃槑 馃摲
Original photo
generated with Photoshop Mehdi Kaleidoscope Filter
generated with Paint.Net Distortion DPL Kaleidoscope Effect
generated with Grok (X)
generated with Qwen (Ali Baba)
generated with Gemini (Google)
Qwen was hallucinating again :) 馃槀
Gemini added some graphics :) 馃榾
馃摲 馃槑 馃摲
Original photo
generated with Photoshop Mehdi Kaleidoscope Filter
generated with Paint.Net Distortion DPL Kaleidoscope Effect
generated with Grok (X)
generated with Qwen (Ali Baba)
generated with Gemini (Google)
Qwen has completely flipped out :)
Maybe he/she/it took it too seriously... Or did take the path of least resistance?
馃摲 馃槑 馃摲
Original photo
generated with Photoshop Mehdi Kaleidoscope Filter
generated with Paint.Net Distortion DPL Kaleidoscope Effect
generated with Grok (X)
generated with Gemini (Google)
Qwen isn't here because I exceeded the daily token limit.
馃摲 馃槑 馃摲
I think that's enough :) 馃槅
To generalize quickly, this is, of course, my personal opinion, because I have generated thousands of kaleidoscopes from photos so far:
I still think I generate better kaleidoscopes in graphics programs, perhaps because I can preview and quickly change parameters, colors, and the like. For AI-generated kaleidoscopes, I need more precise prompts and more time, and I use a lot of resources for results that aren't necessarily what I need.
However, when I compare them to the results from six months ago, the progress is evident. They are getting better and better.
In essence, general LLMs have learned to use graphical AIs. For example, Grok uses Flux1[dev], Qwen uses Tongyi Wanxiang, a multimodal art model via Ali Baba Cloud, and Gemini uses Nano Banana. But I still find that the graphics programs I have installed on my PC offer much more freedom when it comes to designing or changing settings and parameters. And everything runs very quickly.
But it's also true that I only used the free versions. Maybe the Pro versions generate 100 or more kaleidoscopes at once, and then I can just choose. I'll see next time I feel like making kaleidoscopes :)
The original photos are owned by
Maybe this will be useful for 
The introductory image is mine, a kaleidoscope generated with Paint.Net Distortion DPL Kaleidoscope Effect, and the original is my collage for the LMAC Community contest. 馃憞
By the way, I chose this collage for the 1st image because I think it is relevant to geopolitics, right now.馃槑
Thank you for your attention!
Support and vote for those exceptional witnesses:
@good-karma, @liotes, @fbslo, detlev.witness, and @stresskiller. They appreciate any help you can provide.
馃摲 馃槑 馃摲
Stay Healthy!
NFT Showroom
3Speak Channel
Let's Make A Collage
Beer Tasting
New Tunes
Foodies
Ready for Web3
