talk2html

23 Nov 2025

I like a live event where the format is people riffing, and try to attend these when I can. It’s also remarkable that we live in an era with unprecedented access to recordings of these types of live events: I’m always seeing or being sent links to a video recording of a presentation, a panel discussion, or a video version of a podcast.

Unfortunately, the practice of watching or listening to recordings of a talking event has never worked for me: I don’t have a car commute, I don’t use headphones on the train or the bike, and I’m not able to listen to recordings of people talking while I’m on the clock at work. When I do have time at home available for durational media, I’m trying to listen to music, catch a basketball game, watch a movie, or watch braindead TV, in that order.

I do, however, have plenty of other time to read, and this has become even more true as my kid has learned to read and we’ve unlocked “everyone in the family is reading at the same time”. I’d always wondered whether I might be able to catch more of these live event recordings if I could timeshift them from durational media to written media.

To gather some data on this hypothesis, I made a command line tool called talk2html. This tool wraps the excellent yt-dlp command line tool to download videos from YouTube, and post-process them as a filmstrip of images from the video with text presented along-side the recording’s images.

I’m not yet sure if “reading” a talk/panel/discussion/podcast instead of watching it loses something in translation. I just tried “reading” this great conversation between Barbara Fields and Ta-Nehisi Coates; it was cool to be able to read the whole thing where I almost certainly would not have otherwise had time to watch it at all. I suspect this tool will help me to move recordings like these off of my “to watch” list.

But I still found myself jumping to specific parts of the recording where the automatic transcript wasn’t good, or where it was clear that some meaning was being lost as written vs. as spoken. Which is maybe just to say: I do not recommend talk2html for use by grindset guys who are looking to try to pound media (written or durational) like nutraceuticals in a functional smoothie.

On the other hand, I did use talk2html to “read” a conference talk about local-first software, which I won’t name. That talk turned out to be kind of low information and vaporware to my taste, and I was glad to not have spent an hour with it.