We had software that was indecipherable. Then we had software that was possible to use by people motivated enough to keep trying. Then we had software that was fairly easy to use for most people. It’s time to have software that knows how you perform a task and make it easier for you to do it.
For the longest time, I was stuck in the literalist mindset that you’d have to have an AI to do this. No, you simply have to have an application where the developer has built some features around how actual use looks like.
Let’s take podcast apps.
An easy example of this kind of feature for a podcast app is the rewind/fast forward 30 seconds button. It makes no sense for music, which is what was mostly played in media players, but it does for nearly every other type of media, including spoken-word like podcast and radio. A quick setting to speed up and slow down also comes in handy for podcasts.
Don’t get me wrong, I would like to nominate whoever first implemented either of those features for some sort of Nobel prize. But every time someone suggests that this is the height of visionary contextual problem solving, or even worse, that podcast apps are now “complete”, I want to strangle them slowly with a rusty chain.
So threats of (cartoon) violence aside, do I actually have any ideas of this sort? Sure.
Detect the intro theme! This is the foundation of a lot of potential features. It’s almost always identical and almost always completely different characteristically than the rest of the audio and while detecting it will be best, if you ask me to fence it off the first time I play a certain podcast, I’ll do it. The intro theme is often louder than the actual talk and can be obnoxious even in good podcasts. Condensing or skipping the intro theme gives you more time to listen to the actual episode. Knowing what the intro theme is also makes it possible to apply equalization only on the talk parts. (Bad equalization means you boost the volume all the way up and then your ears explode on the intro theme on the next episode.)
Detect sections or bumpers. If a segment regularly appears with an intro or the podcast uses bumpers (small audio ditties to signal a cut or change in subject), you can skip around more easily.
Allow notes, markers or tags. Sometimes I listen to a podcast and there’s a really funny or poignant segment or some odd fact that I’d like to look up later. Yes, I can jump to a notes application and write down episode number and timecode – I’ve never done it. Yes, in some apps I can mark an episode as a favorite – I’ve never done it, and I think the reason is that it doesn’t let me say why this is a favorite.
Invent a standard way to attach links, pictures and notes to podcasts. There are solutions to this, but Apple’s chapter marks means you have to use AAC and most other solutions are inextricably linked with their own apps. Podcasting is the combination of a number of standards and it’s a shame that there’s no standard for these annotations too.
I’m not a podcast expert. That’s the point. I’m just someone listening to podcasts and this is what I’d want. I also know that some of these things are harder to implement than other. Sometimes hard things are hard, and sometimes it’ll be worth it. Computer vision used to take supercomputers. Now it takes OpenCV. Chances are even the truly hard things are not as challenging as I think.