Pixailated: How a Silly AI Experiment Became a Game

About a year ago, I was playing around with AI image generation and had this idea for a game. The concept was pretty simple: take an AI-generated image, pixelate it, and then slowly depixelate it as players guessed letters correctly. Kind of like a mix between hangman and a puzzle. I called it “Nonsense” because, well, the AI-generated images were complete nonsense.

The First Version: “Picsailated”

To get the idea out of my head, I used Microsoft’s image generator to create some random, nonsensical images. I threw together a basic version of the game and shared it with my Facebook friends. I named it “Picsailated,” playing on the idea of AI pictures that were pixelated. It was super basic, with each refresh pulling from one of five random sentences, but it was fun to see how people reacted.

Checking Back In with AI

Fast forward to recently, and I decided to check back in on AI image generation. I was impressed with how far things have come—Stable Diffusion 3.5 is incredible. I got it running on my desktop with my RX7900 XTX, and with its more permissive license, I figured I’d revisit the project.

This time, I shifted the focus to well-known phrases and sayings. The twist is that the images are based on their literal meanings instead of their metaphorical ones. For example, imagine “bite the bullet” as, well, someone literally biting a bullet. I renamed the game to “Pixailated” and released it at https://pixailated.com/ .

Why AI Matters Here

AI has been central to this entire project. The idea itself came from the nonsensical things AI generates, and the game wouldn’t exist without AI-created images. I know opinions on AI can vary a lot, so I think it’s only fair to be upfront about its role in the game. It’s a key part of how Pixailated came to life.

Check it out and see how many phrases you can guess as the images depixelate. Hope you enjoy it!