Users are requesting AI to recreate the same image repeatedly, with each version diverging more from the original.
The outcomes can be both humorous and alarming. In some instances, the images transform into bizarre abstract shapes, while in others, facial features become exaggerated to an extreme degree.
One of the most shared examples is a series of 101 renditions of actor Dwayne “The Rock” Johnson.
someone on Reddit asked ChatGPT to duplicate an image of The Rock 100 times without changing anything pic.twitter.com/IcRgmWHCWK
— Kristi Yamaguccimane (@TheWapplehouse) May 3, 2025
The initial reproductions were fairly close to the source image, but later versions began to warp Johnson’s features and ultimately became completely abstract.
I attempted the “Create an exact replica of this image, don’t change a thing.” challenge. pic.twitter.com/xcAcsvBRJp
— Rish Agarwal (@rish404) April 30, 2025
What’s happening behind the scenes? It primarily stems from the way AI models are trained and the manner in which they encode and reconstruct images.
When tasked with recreating an image, AI doesn’t merely duplicate the original pixels; rather, it deconstructs the image into a complex array of features and patterns that it attempts to piece back together based on its understanding of what the image should convey.
However, this reconstruction process is flawed, which leads to minor inaccuracies or alterations each time the image is processed anew. As the image cycles back into the AI, these discrepancies accumulate, resulting in more distorted or unexpected appearances.
It’s akin to a visual game of “telephone,” where each whisper introduces new variations.
Moreover, the quirks of AI may reflect the biases and assumptions inherent in these models. Some images display exaggerated facial features or adopt a warmer, orange-hued color scheme.
Users have observed that eyebrows often appear overly accentuated—almost as if influenced by social media filters. The prevalence of the orange hue has led some to theorize that warmer tones are favored in photography, making them more prominent in the training datasets.
However, these remain mere hypotheses. The complexities of the vast “black box” that characterizes today’s most advanced models are still largely unknown.
In the meantime, social media users are enjoying the bizarre and occasionally unsettling results produced by this recursive AI image generation.
Trends featuring AI, such as the recent craze surrounding AI action figures, have gained significant traction on social platforms, drawing in thousands of participants on X, Reddit, TikTok, and Instagram.
One user humorously remarked, “I drained the ocean replicating my image 100 times.” While amusing, it also highlights a valid concern.
