1. Idea and Core Mechanics

In my game, the main goal is to catch “snowballs” that drop into a panoramic scene while you move your phone around. I first came up with the idea of combining motion (via the device’s sensors) and a simple winter-themed mechanic. On launch, the snowballs immediately start flying your way, and you rotate your device to catch them. Meanwhile, atmospheric music—composed by a neural network—plays in the background, setting the cozy winter mood.

A simple and fun game to lift your mood and brighten your day!

2. Unexpected Hurdles

As development went on, I encountered unexpected issues. Snowballs stubbornly “hung around” in the scene, consuming memory and overstaying their welcome. The music sometimes continued even after you minimized the app—like an overly enthusiastic caroler in your pocket. I needed to methodically halt the audio engine and implement thorough “cleanup” logic for objects when exiting.

3. Neural Networks Everywhere

It’s not just the music that neural nets helped create. Another neural model was responsible for cutting out a hand image from its background, giving a crisp, lively element to the game’s interface. You could say the entire environment—graphics, audio, even certain code patterns—came together with help from various neural networks. And honestly, I feel like a full-fledged prompt engineer at this point, orchestrating all these AI-driven components.

4. Bringing It All Together

Despite the bumps—panoramas spinning too abruptly or the game momentarily freezing—each challenge led to a more polished result. Now, everything feels in harmony: snowballs fall exactly as intended, timers behave, the neural-network-composed melodies stop when the app leaves the foreground (finally!), and players can simply enjoy the gentle thrill of a frosty catch.