Hacker News new | ask | show | jobs
by crazygringo 2249 days ago
The video on their landing page does a vastly better job explaining how it works:

https://www.grain.co/

I've got to say, that for meetings big/important enough where you've got someone dedicated to taking notes/minutes, this seems like it could be a really effective and compelling way of sharing meeting highlights/takeaways/fun bits with a broader audience, people who couldn't make it, or for posterity. A serious win for workplace culture and communication generally. (Of course, you never know until you actually try using it for a few weeks.)

But if this works well, hopefully Zoom buys them, and we'll expect to see Google and Microsoft build their own versions.

2 comments

Thanks, the video took a surprising amount of work but happy w/ the result.

You're exactly right wrt the dedicated real-time annotator. We created a information transfer framework for our internal product development/design that has helped us to see that for our product to work (highlight clips that don't suck)... you need a dedicated person who's job it is to annotate the content of the video stream in real time (aka taking notes). It's too complex to do well with AI... let the humans drive the car for a while if you know what I mean :)

Our core use cases so far have focused on where this dedicated notetaker is already the case and over time we can automate the annotation process and expand to more use cases without a dedicated notetaker.

The cool thing we've seen since releasing the ability to take collaborative notes w/ your team is that the annotating just kinda happens naturally... people jot down things they thing are important to the mutual benefit of everyone on the team (in the meeting and not).

I love that the automatic transcription shown in that video has errors. It’s realistic, while also showing that enough is accurate that it’s useful.
Yup, you nailed it.

Our transcription will naturally get better over time but it doesn't have to be perfect b/c it's a means to an end in Grain. (creating/trimming video clips)

Even perfect transcripts are still terrible to read most of the time b/c the dynamic nature of spoken dialogue just doesn't translate well across mediums.

Google "Media Richness Theory" if you want to go down the rabbit hole on this.