Any place to get the segment timestamps for TWIT and Sec Now shows?

Ok, @Leo went through TWiG and did timestamps in the discussion thread…

It is a lot of work…

2 Likes

@big_D Wonder if it’s possible to listen to a podcast with a phone that has a speech to text (S2T) app running. When it’s time to record a timestamp (aka time code), you pause the podcast (or video) and say into the S2T app, “12:36. Breaking news; Linus Torvalds accepts offer to become the next Microsoft CEO”.

Then at the end of the podcast (or video) you have a text file with with the times and segment titles. (Which is all I need.) That text then gets included as part of the show notes bundled with the uploaded podcast file.

Side note: Given how AI is developing, in a few years you’ll just upload the audio to an online service and a large language model AI will accurately do all this for little or no cost. Or it will be a standard feature of podcast hosting services.

#TWITArmy4timestamps :wink: :grin:

:sweat_smile: made me think of the old Top Gear UK gag phrase “How hard can it be?”

2 Likes

Thanks for doing that Big_D! It is a lot of work. We’d have to add several hours of staff time per show; I’m reluctant to call for volunteers. Honestly, it’s just a luxury we can’t afford right now. But I will bring it up at Editorial tomorrow.

5 Likes

Many kudos and thanks to you, Leo, for your responsiveness and hard work creating so many great shows and fostering a great community. :star_struck: :blush: :tada: :pray:

1 Like

Update, I sent this message some some NLP AI companies:


I’d like to see an AI app that can listen to the audio file of an English language podcast, determine when there’s been a change in topic and note the timestamp time code when the change occurred.

I listen to various tech podcasts. A single podcast episode will cover a number of topics. When I email the podcast creator asking them to include timestamps for the various segments they say it’s too much work. So, in general, I don’t subscribe to podcasts that don’t have timestamps, even though they have good content. I don’t have the time to manually hunt down just the segments I want to hear.

I bet one of your Large Language AI Models could solve this problem and automate this task. I’d be super happy to see this solution be made available to podcasters. Make it so.

Maybe integrate the solution as a feature provided by the big podcast hosting platforms. Thanks!


I listen, occasionally, to TWIG and only occasionally because the shows have grown so long. When I stated listening, in the days of Gina and Matt and Leo, the shows were 90 minutes. A rare, long, show might get toward 2 hours. Now, a typical show is over 2 1/2 hours and frequently reaching almost 3 hours. I will check the show notes to see if there is a segment I would like to hear but usually don’t even listen because it is too much trouble finding the segment. I don’t know that timestamps for each change of topic are necessary, but maybe 4 or 5 to divide the show into “chunks”. If you want to listen to a segment in chunk 2, you could search within it. It may not meet the needs of people who want every segment timestamped, but I think it would help a lot of people.

1 Like

This is exactly my issue. There are many good podcasts I don’t subscribe to (like TWIT and Sec Now) cuz I’m unable to skip the segments I’m not interested in.

From what I can see, dividing a show into “chunks” wouldn’t work for me. I’d still have to spend frustrating time going back and forth finding the segment I want. That’s a show-stopper for me.

As audience demand for timestamps grows, they’ll make more tools to automate the process. As the tools get better, creating timestamps will (hopefully) soon become a one or two click process.

I’d be happy with just the plain text of the time codes and the segment title. If there’s no clickable time codes or chapters display, that’s fine with me.

I use Otter when researching a podcast episode.

Also Descript has potential.

Both companies should be advertising on TWIT.

2 Likes

I looked at the links and didn’t see either mention AI recognition of different topic segments (and time codes) in a podcast audio file. That’s the key technology I’m hoping for. Aka AI recognition of when the topic changes.

Personally I wouldn’t mind if the time stamps were just for the big topic changes; it doesn’t even have to be the minor discussion changes within a single topic. I don’t need to see going from one individual feature of some story to another, just going from one story to the next. Just the bullet list of topics on the show notes to have time stamps. Although a second view of the show notes shows A LOT more bullets than I initially thought was there, so this would still be a big task. But if there’s just 8 stories, just 8 time stamps, but again, FAR easier said than done.

2 Likes

Yep, this is also what I want. Hopefully some AI company will soon make it very easy to do this. From what I understand, current AI systems already have the ability to easily do this. They just need some training on good data (for example, the 700 plus Sec Now transcripts would be a great training data set).

It seems like YouTube is auto-generating timestamps for some podcasts. I only noticed it for some of the shorter episodes, like Hands-On Tech and Hands-On Mac.
image

Edit: I tried asking ChatGPT to generate timestamps using a transcript. It did pick up on some of the topics, but the times were always incorrect and incomplete, so it’s not really usable. At least with the keywords I tried (“Create timestamps of topic changes using the following transcript: pasted entire transcript”)

1 Like

Very cool. Thanks for the update. Like almost everything in AI, it will quickly improve. If there was some way users could easily edit and correct the YT AI suggested timestamps, that would (I think) really speed up the training process.

I’m guessing YT wouldn’t rollout the auto timestamp feature unless they plan to take it all the way to high accuracy. Could be a way to entice more people into a YT premium sub.

There’s lots of good podcasts I don’t listen to (like TWIT and SN) simply because there’s no timestamps. I’m eagerly awaiting accurate auto-generated timestamps for podcasts! :grinning: Hopefully that’ll become a standard feature on podcast hosting platforms.

Not sure about training, but in general, the owner of the YouTube channel can edit the time stamps that are displayed in the video (description and position) similar to subtitles.

A different approach would be manual timestamps. I believe the WAN Show just takes the time stamps from a YouTube comment and credits the author in the description. When TWiT still had YouTube comments enabled, I saw people would sometimes post time stamps of the show.

But YouTube isn’t TWiT’s primary platform, so I don’t think we will see a change here.

I’d love to do it but frankly we don’t have the staff to do so. If someone wants to volunteer maybe that would work.

Every time something like this comes along it’s non-standard (in other words locked into a single platform). If there ever is a universal way to do this we’d be interested for sure.

1 Like

Leo, is the YT channel of Ask The Tech Guys set to children friendly, I always wondered why we are unable to chat in the comments, like those children YT channels. Also there are times when you would find a dedicated listener going through and making timestamps for the entire episode, seen it happen for many podcasts, but without the comments it’s not possible.

1 Like

I’m not @Leo and my opinion carries no official weight… but YT comments are more trouble than they’re worth, especially since Leo has said they make very little from their content on YT. IMHO, too often you see scammers and griefers and other mal-contents on YT comments… to the point that I refuse to ever look at them.

That’s right. We don’t have the bandwidth to moderate YouTube comments, and they definitely need to be moderated. We have many places which are monitored, like these Forums, where you can comment.

1 Like

From what I remember, some of the YouTube comments used to be very short and negative towards the shows or the hosts with little facts. This is a problem on YouTube in general though.

However, other YouTube comments were more polite and correcting some incorrect info or providing additional context, so I think they had some value.