Always love to see new baseball visualizations. Not necessarily a big fan of AI art, but it's cool how dynamic it is. Some constructive criticism: I think using a real pixel font and maybe writing a deterministic downsampling algo for the images instead of relying on AI would go a long way to make this look better.
Not to hijack your thread but in case anyone's interested in a physical scoreboard built on top of the same APIs using Raspberry Pis, I have a project as well. We also support software emulation if you don't want to buy parts.
One of the side projects I work on is a scoreboard that displays MLB scores. It's highly configurable -- you buy the size panel you want and a Raspberry Pi, install the software, and you can configure it to display games, standings, and news headlines for your favorite team or division.
The problem is that the hardware is purchased by the end user, so it can come in many different sizes. I think we officially support 6 or 7 sizes right now, and each panel can be a chunk of change if you get a nice one. If we wanted to test on every device that means I need to shell out 50 bucks x 7 sizes, plus Raspberry Pi and wiring adapter, so not insignificant for a hobby project. Instead, I wrote a drop-in replacement emulator that makes it super simple to emulate any size panel across a variety of display types.
The most advanced display adapter spins up a minimal webserver and serves emulated images over a websocket, meaning you can display your panel over the network on pretty much any device with a web browser.