Happy Devember 2020!
Everyone loves the L1T news! Those rascals have a running joke giving "protips" for people unfortunate enough to make the weekly news. It would be fun to see a supercut of all protips from the history of the show...but finding them all?
That's what these scripts do! They use the Youtube auto-generated subtitles to locate the timestamp(s) of the phrase "protip" inside every video and download the clip!
<iframe style="position:relative;top:0;left:0;width:100%;height:calc(90vw/1.777778);max-height: 395px;min-height: 255px;" src="https://www.youtube.com/embed/_K_5wVtvGqA" frameborder="0" allow="accelerometer; autoplay; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>Ok, to get this done you'll need a bash shell with the following tools.
youtube-dl
ffmpeg
jq
python3.6+
For youtube-dl
and ffmpeg
whatever version is in your package manager - it's out of date. Go directly to the project website to download the latest version!
Cawling for "protips" is completed in three steps:
-
Start by cloning the repository, it looks like this:
$ tree -L 1 . ├── subs ├── clips ├── tips ├── download-protips.py ├── download-subtitles.sh └── README.md
-
Download auto-translated subtitles for every episode of the L1T news into the
subs
folder. This script uses the youtube playlists for every "season" of the news - for future seasons, add the playlist urls.$ bash download-subtitles.sh
-
Search the subtitles downloaded in the previous step for the phrase "protips" and download clips at that timestamp. The clips will be saved into the
clips
folder.$ python download-protips.py
This process involves downloading every episode with a protip so it's going to take a while and use a ton of bandwidth. I've hardcoded the limit to download the videos at
<=1080p
resolution, which speeds things up but it still takes hours. If your internet is slow (like mine) consider renting a VPC with gigabit internet and running it there!
At the end, you're left with a tips
folder full of short video clips of all found the protips!
The scripts certially work, but I can't be sure they find every last protip. I don't know if you can really be sure of the count without re-watching the entirity of the news. However, it does work well enough to make a fun supercut like this!
If you think you can improve the match-rate, feel free to fork and make improvements. Happy Devember 2020!