vtt-compactor: slightly less verbose VTT files.

I’m working on a small project that involves interviewing a number of folks to capture their experience becoming and working as a Staff engineer. Hopefully the first stories will be ready to post in a couple of weeks, but what I’ve been surprised by is just how tricky it can be to get good interview notes.

So far Zoom’s record meeting has been phenomenal because it produces a text transcript in the VTT format that’s extremely “good enough” to edit into an interview. However, the format is a bit verbose because it’s meant to be used for subtitling:

24
00:03:46.020 --> 00:03:53.970
Will Larson: That's amazing. I have so many questions that are
off script that I want to ask on but like, how do you measure
that a lot of people feel about it like that. I want to ask, there's

25
00:03:54.420 --> 00:03:58.560
Will Larson: I think the fact that somebody's going to be using program.
And I've actually never heard anything quite like it before.

This meant that I was spending a lot of time manually removing the metadata before I could start on the more important work of editing the content. To save my sanity, I wrote up a very simple Pythons script, vtt_compactor which you run like:

python3 compact.py yourfile.rtt

And creates output like:

24. Will Larson. 00:03:46.020 -> 00:03:58.560

That's amazing. I have so many questions that are off script that
I want to ask on but like, how do you measure that a lot of people
feel about it like that. I want to ask, there's
I think the fact that somebody's going to be using program. And I've
actually never heard anything quite like it before.

This is very small improvement, but has eliminated about thirty minutes of manual editing time from each transcript so decided to share it out. (To be honest, would probably take closer to ten minutes to do it if I focused, but it’s sufficiently annoying that I find that I keep distracting myself and it takes longer than it ought to.)