Abacaba | The history of the most viewed YouTube videos, converted into sound (2019 re-upload) @Abacaba | Uploaded 5 years ago | Updated 3 hours ago
Original upload date: Aug 20, 2017.

Behind-the-scenes video (6 hours of me coding): youtube.com/watch?v=kRK8E6gF_DI

As it says in the intro:

A few months ago, this entire channel was demonetized for "reusing content". All the videos on this channel are original data visualizations created with code I've written myself, so I wasn't sure what the cause of this was at first. After some investigation, I realized the problem might be my "YouTube choir" video. In this video, I played 5-sec chunks of popular music videos to demonstrate how I could pitch-bend them to the frequency at which they got views. I believe this is fair use, but I don't want to contest it and risk more punishment.

Now, I would like to get this channel re-monetized. I don't care about the money, but I do know YouTube suppresses the spread of demonetized videos since they can't make money off of them. I do care about that, because if I make a video with a message about measles, I'd like as many people to see it as possible.

So, I'm reuploading this video with the problematic sections removed. It was the most popular video on this channel, so it's unfortunate that the views, comments, etc. all start at zero again. However, it's much better to kill one video to save the channel. So here is the reupload!

**** OLD DESCRIPTION ****

Why did I choose the videos that I did? Well, the original plan was to include only videos that have reached #1 at some point in time, which is mostly music videos. Including other music videos that haven't gotten to #1 (like Hello, Sorry, Uptown Funk, etc.) isn't such a good idea. Since they constantly get views, we'd quickly get clouded with dozens of voices singing at once, which'll turn into a muffled mess. That's why flash-in-the-pan viral videos are so great - they add interesting events to the video, but don't cloud up the audio space for more than a few seconds! My other criterion was a little subjective: it had to get enough views to reach a comfortable pitch (about 2M views a day), but it also had to be considered part of "YouTube culture", like the ALS ice bucket challenge certainly was. So I didn't include Adele's carpool karaoke since that seems more like traditional TV.

There is very little data from before mid-2009, so that's why I didn't go back that far.

To get the majority of the data, I used inspect element to find the coordinates of the vertices of the "daily views" graph underneath each video, and converted those coordinates into views.

For the videos whose stats were private, I could usually look up the SocialBlade stats of the channel that uploaded the video. If the main video accounted for over 75% of the views, I could somewhat safely just assume that that fraction stayed about constant over time. So I just multiplied the channel's total upload views by that fraction, and voila.
socialblade.com

For videos that were older than SocialBlade itself, I used this Wikipedia article (en.wikipedia.org/wiki/List_of_most_viewed_YouTube_videos) for some historical data, and then interpolated between those given data points. To be extra fancy, I didn't just linearly interpolate. Instead, I looked at the Google Trends (which goes back to 2004, before YouTube) for that video title, and assumed that the number of searches was somewhat proportional to views received on that day, and interpolated using that. All that work probably wasn't worth it, though, because if you look at the two oldest videos (Evolution of Dance and Charlie Bit My finger) their view-graph lines still look pretty close to linear to me.

Music:
Almost Original (Instrumental) by Joakim Karud http://soundcloud.com/joakimkarud
Music provided by Audio Library youtu.be/r20_9c0fzGk