I'm having a hard time understanding what exactly the graphs are showing. Is it just average pixel color per frame?

I think why the shape of movies works is because it's taking the entire video, where each cinematic shot is almost visible because most (all?) frames are included in the generated image, while yours is less than ten frames so there is no patterns emerging that looks visually interesting and pleasing.

Netflix Tech Blog has some really neat stuff on this topic, apologies, idk what to link exactly but was reading it last night and pairs directly

Cool project. I do urge you to add some explanation to the top, though.

Hi everyone! Glad HN gave this a second chance. I made a video explaining all about the site, how I used k-means clustering, threaded workers… It’s here, hope you enjoy it: https://youtu.be/jvFoGdvgWBs

I wonder if you might advise me. I'm presently using yt-dlp with python selenium to expose censorship of comments, analyzing triggers and exposing the multi-universe false presentation of public discourse falsely presented by shadow-banning, url manipulation, mobil vs desktop and cookie or login status.

People are literally seeing different realities of serious (sometimes) discourses. Additionally, Shadow-banned commenters see their comments in the thread but no one else does.

I'm analyzing both yt-side and channel-operator side censorship and its arguable effects on society.

There's more, but my comment will become grayed out soon so not really worth saying more.