I've always enjoyed watching police bodycam videos and finding out that it was close to me, so I was surprised when I found out that nobody had tried to make an explorable map of them yet.

I used yt-dlp to rip bodycam videos from youtube, distil-large-v3 to extract transcripts on my rtx 3070, and deepseek-v4-flash to output classifications. I host the entire thing on GitHub pages with an issue/PR system that allows the community to submit corrections and human reviews for metadata. This way, it's totally free for me to host, aside from the domain.

I think it would be cool to expose some more information later down the line (you can see in the dataset that I have already extracted more than what is shown).