Hacker News Reader

Comment by leobg | Hacker News Reader

Wish there was a service where I can plug in the URL of this podcast to get a Whisper transcript, separated into paragraphs, and with headlines for skimming. So I can decide whether or not to invest an hour of my time to actually listen to the whole thing.

View on HN

There's already clues there as to what sort of angle this might be approached from. The speakers book "Jerks at Work: Toxic Coworkers and What to Do About Them, by Tessa West, 2022." as well as their work and other cited research is listed on the page.

I just started listening, there were segments with the first being 15m. I would say that they're a collection of anecdata stories on the subject. I napped for a segment in the middle, dreaming about visiting a treehouse office, and was awake again for the end. I didn't find any big answers but also don't mind having passed the time.

I'd rather get your takes on episodes as a paid service over the idea of the parent :p

I agree -- where can I sign up? :)

So you'd pay someone to listen to Podcasts and send you a summary?

But what would we call it? Slashpod? Poddot? Either way, you'd be continuing the proud tradition of not reading the summary.

The "meat" of the thing was right in the middle, so you may have missed it... it's simply that according to her research the best way of figuring out what someone is thinking or feeling is to ask them. Doh.

Thanks, this is one of those times where it would have been more useful to read the comments first.

On a positive note, the voices were both soothing enough that I did get a short refreshing nap out of it.

I wonder how many podcasts are actually being abused as sleeping aid.

Transcribed using 'base' model:

https://pastebin.com/M5276AFE

Wow. Thank you!

I'm surprised by how good the result is, what tool did you use for this?

Judging from OP's comment, they used OpenAI's Whisper speech-to-text transcription model, which was recently released open-source: https://openai.com/blog/whisper/

Ah okay, thank you, I was not aware of this and so I didn't know what to search for. Thanks a lot for the help, much appreciated.

No problem! It's getting harder and harder to keep up with all the stuff getting released in the AI world these days.

And a stab at auto-summary using tools4noobs:

https://pastebin.com/VY6M489T

You can search here:

https://www.happyscribe.com/public

Didn't find Hidden Brain through a quick search, maybe it's there and I didn't have enough patience.

I'd love to know if there are other services similar to happyscribe.com .. anyone know any?

It adds so much value to a podcast to be able to search for words/text and read instead of just listening.

Also, if you want to add your own thoughts to something that was said, with a transcription you can simply make a PDF and attach an annotation or highlight sentences.

If like me, you like to listen to podcasts and take notes and add your own thoughts, and if you have found some handy tools for that or come up with your custom convenient workflows, I'd really like to know.

Thanks for the recommendation. Didn't know that service.

Why are you asking for similar services? What's something that you feel is missing on this one? If you could change it or add functionality, what would you change or add?

They only have a selection of all podcasts.

I once used their trial tool, to transcribe some episode. The functionality is good, but it is commercial. I wouldn't use it often enough to justify signing up.

So if there were some similar tool where I don't have to sign up to some paid service just to transcribe one episode here or there, that'd be neat.

Read of Whisper in the comments. I like that it doesn't appear to require some API keys to use their model. But I'd rather not install yet another Python package that likely comes with a whole bunch of new dependencies.

You can try my C/C++ port of Whisper:

https://github.com/ggerganov/whisper.cpp

No dependencies, no Python, runs efficiently on the CPU.

Thanks, does indeed look neat.

the answer is "You ask them". Saved you one hour.

slaps forehead

Why didn’t I think of that before?

Better yet.. podcast -> Whisper -> text summarization model... sounds like it'd be fun to hack together

If you do, please post!

whisper.sh:

    $ curl https://news.ycombinator.com/item?id=33052985 | grep -A10 'tldr' 

    jderick 4 hours ago | prev | next [–]

    tldr:

    "So you're saying the best way of figuring out what's inside someone else's mind is to ask them?"

    "That's right."

I made a similar thing for the Lex Friedman podcast. It transcribes the audio and runs it through a summarization model. You can then search the entire corpus to find relevant snippets from episodes:

http://lexicon.lol

FYI .. all episodes are available transcribed here: https://www.happyscribe.com/public/lex-fridman-podcast-artif...

what's a Whisper transcript? (looking, enticed)

Transcript created by OpenAI's new "Whisper" language model.

thanks

Reply on news.ycombinator.com