The concept was simple. Take a few of my favorite podcasts and run them through automated speech-to-text and make the result searchable.
It works. I’m still waiting for Google to add real contextual search to video and audio. They’ve got images working well. And Pinterest has even taken that a step further. But, at some point, every bit of content should be searchable.
This reminds me of a tool I wanted to add to Viddler years ago. I described it in this blog post. I wrote:
I remember in 2008 or 2009 when I was working at Viddler I had come up with a conceptual way of pulling this off for our platform. We never fully implemented it. But I did take a swing. I still have the code.
It went something like this; every video has a certain number of keyframes in it. You can think of those keyframes as thumbnails. In fact, at Viddler we stored several of those thumbnails per video. Imagine tagging someone’s face in a video and using facial recognition on the rest of the keyframes just to mark where in that video the person was. (at the time, face.com’s API was still a thing, it could have been done for free).
Imagine if that existed? Seeing _DavidSmith’s new side project makes me want to build it.