TL;DR - I hate ads and built a thing to make them suck less in sports broadcasts.

I am, unfortunately, a big college football fan. Not even just my team (Go Blue!), but like, the whole sport. In other words, I’m a college football sicko. The biggest downside to this lifestyle is that I have to put up with a deluge of ads every Saturday. I’ve found it especially vexing now that it seems like half the ads are trying to get me to gamble on the games.

In light of this I started recording games via YouTube TV and sitting down to watch them about an hour and a half after kickoff. This allows me to skip through commercials and halftime and catch up to the live broadcast somewhere in the fourth quarter. I’d always wondered whether it would be possible to build a system that can do this for me automatically, and it turned out it was!

With the help of Claude I put together a system that allows me to automatically skip ahead when a commercial or halftime is detected. The basic setup is an Apple TV with an HDMI splitter - one split goes to the TV like normal, the other runs to a capture card connected to a little linux box. I build a little profile for the game by selecting the logo in a sample frame, and building a color histogram based on ~30s of gameplay. The box checks frames against this profile, and if it’s too dissimilar, it sends the “skip” signal to the Apple TV via pyatv. There’s a little more to it than that (debouncing, in/out thresholds, etc), but that’s the gist! I’ve also included an option to simply mute the TV rather than skip if you’re watching the live broadcast.

The setup is surprisingly effective! Being a data scientist, I of course tried a few more ML-y solutions, but the simple threshold version is much more intuitive and tunable. Football season hasn’t started yet, so I’ve only used the system to watch the NBA playoffs. It worked beautifully! I’m excited to have a slightly less DraftKings fall.