devtesla wrote on March 17, 2025 at 5:05 AM

I have two things to regretfully inform you of:

1: I occasionally listen to the podcast of the guy who makes the podcast app I use (overcast) and he thought up this idea a decade ago. He never implemented it because he didn't want to fuck with people's business, at least back when the whole thing was smaller scale. No idea why I remember this random thing, I guess I also thought it was a good idea.

2: The way ads are delivered today would screw this up. Podcasts (or at least the one with lots of ads) use dynamically inserted ads now, meaning that there isn't just one file that everyone is downloading that is the podcast. You can download a podcast on Monday, and then download it again on Tuesday, and it'll have a different number and length of ads. There's marketing tech in the background adjusting things based on what people are paying for campaigns and such. And while they don't track that much about you, they can tell generally where the file is being downloaded from a target ads that way, so two people downloading the same podcast at the same time might get different ads.

I also learned this from that podcast app dude. He had to remove a feature where you could stream a podcast from the internet, pause it, and then resume it without downloading the whole file to your phone. It would try to resume from the same time but it'd be jumbled around because it wasn't the same file anymore!

Ads fucking suck I hate them so much!!!!!

flabberghaster OP wrote on March 17, 2025 at 5:42 AM (edited on March 17, 2025 at 5:46 AM)

The dynamic ad thing I knew about but I didn't think of it when I made this. The "if it got big it would crater the already not great podcast ads business model and put people whose output I like out of work" did occur to me.

I feel like you could may be work around the dynamic ads issue if you were good enough with signal processing. Originally the idea was "the only good use case for an AI I can think of, is one trained to recognize ads in podcasts and skip over them" but then I tried to think how to do it without AI.

Even if you didn't use AI, you could probably do what Shazam did where it could recognize a song. The app would instead, recognize the point in the podcast where they were just about to throw to ads, and when they come back, and skip over anything in between. This would still work with tagging probably.

The podcast companies would come up with countermeasures I'm sure. And it seems quite complex. But I wouldn't have to hear about sports betting as much.

devtesla wrote on March 17, 2025 at 7:00 AM

Sports betting ads is an instant unfollow for any podcast Lol.

What do you listen to anyway? My regulars are Just King Things (Stephen King podcast, weirdly touches on a lot of American culture), QAA (conspiracy skeptic stuff), some board game stuff, and weirdo comedy stuff like Seeking Derangements and Doighboys.

flabberghaster OP wrote on March 17, 2025 at 4:32 PM

I listen to a ton most of which have no ads, but Behind the Bastards, It Could Happen Here and Better Offline are all I heart radio and they have all kinds of fuck ass ads.

It's so funny to be listening to better offline, where the whole podcast is about how bad AI is, and every ad is about how you NEED to up your AI game to keep up in the market.

neku wrote on March 17, 2025 at 7:11 AM

couldnt you circumvent dynamic ad insertion by like, developing a way to recognise ad audio data like how normal files can be recognised by md5 checksums? for each podcast users could flag ads and upload their like, "audio signatures" to a central server. when users play a podcast they could download the audio signatures and automatically skip periods of a file which match them