Hacker News new | ask | show | jobs
by corobo 3835 days ago
We currently don't have the technology to automatically transcribe podcasts into text (good for the SEO on your podcast site) - But even if we did, how would it know that something is an advert as opposed to part of the show? Most of the podcasts I listen to have ads and in many cases the ads aren't just at the start or end, they're either a live read during the podcast or a cutaway somewhere towards the middle.

Maybe you're just not the target market for ad-funded podcasts.
1 comments
scottjad 3835 days ago
Cut any segment of 10 seconds or more that is found in more than one podcast episode. (I think this is feasible considering Shazam)

It would remove pre-recorded/repeated ads in the middle (assuming that's what cutaway means), but wouldn't be able to remove live reads. It would also remove intros/exits.

link
corobo 3835 days ago
That actually does sound feasible. I'm not sure how feasible when it comes to mobile devices but definitely possible if things like YouTube's ContentID can work
link
dragonwriter 3835 days ago
As I understand it, ContentID is notoriously over-aggressive in matching and only "works" in the sense that YouTube's interest in having it isn't particularly harmed by that, since its mostly a tool to improve relations and avoid lawsuits from big media interests.

Without manual validation, it probably wouldn't be a good model for identifying and removing ads from podcasts, especially using a "repeated in multiple podcasts" model, which doesn't start with known ads.

link