This not on my short term TODO list, but I think it could be interesting to use the Rapberry Pi power to identify the ads with audio fingerprinting using this lib

Audio fingerprinting

The concept behind audio fingerprinting is that the lib analyze a sound/song and extract some unique characteristics about it. Later it can quickly recognize it. The most famous example is Shazam.

So by fingerprinting the TV jingle announcing ads, you could detect ads automatically and switch to your video clip instead. And since each channel has its own TV jingle you can know which channel the user is watching and the probable length of the ads.

Audio source

And how would I get the TV audio you ask? Well, luckily some TV have a TV audio out plug in case you want to get your TV audio on an amp instead of the TV speakers. If you are lucky, your TV has this audio out and it's analogic. In this case you just have to get a 5€ USB audio card and you can capture audio with your Pi.

If your TV doesn't have such an output, you may try a HDMI splitter with audio like this one.

Well, don't expect to find the code for this feature in the repo this week, but it's definitively something I'll try.