A scene is one thing, the next is a capable photographer, then there is the editor. But thanks to AI, some of the works can be offloaded.
Since AI can create images out of thin air, the internet starts to see a rise of 'Hidden Words in Arts'.
This refers to AI-generated images with optical illusion that spells out words or letters using the figures inside the images.
Often, to see these hidden words, viewers have to step backward, or see a smaller version of the image, or squint their eyes.
These images only began appearing online alongside the rise of AI content, generative AIs, and more.
The first of the images, can be traced back to Reddit, where a user posted a tutorial to create "images with 'hidden' text using Stable Diffusion and ControlNet."
The post quickly received upvotes, counting up to thousands in just a couple of months.
In one of the first viral example, is an image of skyscrapers in a big city, with the sun shining revealing the word "New York."
The Redditor made use of the trend that picked up about a month earlier, during the time when a user uploaded a package to Hugging Face, a platform where AI researchers and enthusiasts post code and models they're working on, that allowed users to generate QR codes hidden inside of other illustrations.
The package worked via ControlNet, a function that is then added onto the AI art generator Stable Diffusion.
ControlNet, which was introduced in a February 2023 paper, allowed users to give the Stable Diffusion two different prompts at the same time.
In this case, the first prompt would be the text that wants to be hidden, and the second prompt is to create the image the text is supposed to be hidden in.
Then, using Stable Diffusion, the AI would produce the two images at the same time.
Soon after the trend picked up traction, a whole bunch of images with hidden words were created and spread through the internet and social media.
"Obey"
"Stop"
"Game"
"Pornhub"
"Bitcoin"