2
1 Comment

Take a glance at any podcast!

Hey,

I have created a system that can extract the highlights from any podcast and join them together to create a ~2 min videos.

Here is the video: https://twitter.com/GlancePod/status/1266133069679013888

I have implemented a modified version of a natural language processing algorithm to extract the context and information on any podcast then pick the highlight sentences from the transcript. Even though the podcast doesn't have a transcript, it is able to convert speech to text and apply the implemented algorithm on it. It is still of course, a work in progress.

Every little piece of the video is automated and I haven't done anything other than running the script to create this video.

Quick question: What do you think? Does this solve a problem of you? If yes, how can it be more useful?

on May 28, 2020
  1. 1

    I think this is super-interesting because everytime someone sends me a podcast to listen to, even though I truly believe people I care about wouldn't share crap with me, my initial reaction is still "OMFG, I don't have time to listen to an hour of this!!" So, the pain point is there for me, for sure - like a sort of Blinkist for podcasts.

    However, this didn't really solve it for me. Other than make me go "I cannot handle GaryV" and "If those are the highlights, I definitely would be wasting time if I listen to the whole thing", which maybe is solving it in a way ;)

    How does it determine what is a highlight? I'm not sure I heard the best bits of that podcast. Because they were extracted from the middle of topical conversations, I didn't have any context (like where they talk about inviting people to apartments to have parties, I'm like "why were you doing that??") So maybe a better way to highlight things?

    Maybe using snippets from the podcast itself is not that useful. Can you generate a more useful summary, and audio-fy that? (This might be magic, I'm not sure it's possible with current tech)

    Good luck!