Mirroring YouTube Playlists

Recently I wanted to set up a periodic job to mirror some of my personal youtube playlists. There’s plenty of reasons one might want to do this. For me it’s simple: one copy is none copy, and two copies is one copy, so I want a second copy of youtube videos I care about stored locally. This protects against videos getting removed, copyright stuck, youtube shutting down, or anything else that might make the youtube video otherwise unavailable.

The final straw for me was the recent YouTube Kids change. YouTube has implemented a system whereby it flags videos as “For Kids”, and a video flagged as “For Kids” can not be added to playlists, or interacted with beyond viewing, and may eventually start disappearing from search results. This seems kinda fine on the surface, but the automated systems flagging videos are very bad at actually detecting kid-friendly content. For example, Pony Music Videos (PMVs), which are fan made cuts of My Little Pony video footage over top unrelated songs, have been getting flagged quite a bit, even when the song has obviously explicit lyrics. (EDIT 2020-03-02: this video has since been un-marked as “For Kids”, so you’ll have to take my word that it was “For Kids” when this post was written.)

The Script

Anyways, I just wanted a simple script adding some extra features around youtube-dl, so I made one! You can find the full script at https://github.com/faithanalog/x/blob/master/youtube-archiver/archive-playlists. I’m going to go over a few specifics in case you’re interested in making your own rather than just using my script or someone else’s. But, if you don’t care about that, and you want to use my script, here’s what to do:

First, create a list of playlists in a file called, for example, playlists.txt

Nightcore Songs


Optionally, create a list of SOCKS proxies you want to use in another file, for example, proxies.txt

Then you can run my script!

# If you don't have any proxies to use
archive-playlists playlists.txt

# If you do have proxies to use
archive-playlists playlists.txt proxies.txt

This will create folders with the names provided by playlists.txt, and download each playlist to its own folder. You can run it periodically however you feel most comfortable (crontab, systemd timers, etc.) and it should keep all the local copies up to date.


The requirements I had for my script were pretty simple:

Most of this can be taken care of with various youtube-dl flags! I have a function called dl_playlist() which implements all of this. Arguments passed to dl_playlist() are transparenltly passed on to any youtube-dl commands, which allows me to pass it a playlist URL, and optionally a proxy.

First, we download the playlist metadata:

playlist_json="playlist-$(date --rfc-3339=date).json"
youtube-dl \
    --flat-playlist \
    -J \
    "$@" < /dev/null > "$playlist_json"

temp_playlist_json="$(mktemp -p ./)"
cp "$playlist_json" "$temp_playlist_json"
mv "$temp_playlist_json" playlist.json

This writes the current metadata to a dated file, copies that to a temporary file, and then renames the temporary file to playlist.json. This allows us to have a history of the playlist over time, which may make it easier to figure out what videos are missing when they get taken down. playlist.json will always contain the latest playlist information, to make it easier for me to write tools for this later.

The process of copying to a temporary file and then renaming the temporary file may seem overkill, but mv is an atomic operation, so it eases my mind about any race conditions I might run into if I’m running other tools over this data to process it.

Next, we can download the playlist videos.

youtube-dl \
    --download-archive ytdl-archive.txt \
    --write-info-json \
    --write-description \
    --write-thumbnail \
    --all-subs \
    -i \
    -f bestvideo+bestaudio \
    -r 500K \
    --sleep-interval $min_sleep \
    --max-sleep-interval $max_sleep \
    "$@" < /dev/null

Here’s where all the fun flags come into play!

The rest of the script is just housekeeping around this. It loads playlists from a file, creates separate directories for each one, and randomly shuffles through a list of SOCKS proxies for each playlist. My VPN provider provides SOCKS proxies for each of their exit nodes, so this is a really convenient way for me to distribute my downloads across a broader range of IP addresses.

And that’s it! The script is pretty well commented I feel, but if you have any questions about it, feel free to ask!

Thanks to everyone on the fediverse who helped me iterate on this script and iron out the details. ❤️

– artemis