I've now expanded this with an additional script, a github repo, and proper documentation.
Here's a cleaner explanation:
I've taken on the challenge of self-hosting more of the services I regularly depend on. The latest target is Spotify. This meant I needed a simple and convenient way to listen to my music from anywhere, get new music into my collection, and also still receive recommendations based on my interests and listening habits.
I now have what I think is the pretty ideal setup, here's what it includes:
A 24/7 radio station that plays my entire catalog (link here if you're interested). This is powered by Azuracast along with the scripts in the repo. The station link is using the Public Pages feature in Azuracast with a bunch of custom CSS.
A Spotify-like experience that also supports mobile and offline. This is powered by Navidrome for web/desktop and Substreamer for mobile. Substreamer connects to Navidrome using the Subsonic API.
A couple of scripts that allow me to easily download tracks/albums/playlists from Spotify and Youtube. I used these to bootstrap the collection and export my existing playlists from each service.
A couple of scripts that automatically grab my latest recommendations from Spotify and LastFM, add them into Navidrome, and provide me a nearly fully automated way to parse out tracks I want to keep permanently.
That last point is the most interesting part in my opinion. Both scripts run on a weekly cron job that downloads my Discover Weekly playlist from spotify, and current recommendations from LastFM. It then creates a playlist for each source for that weeks collection and moves it into Navidrome.
I then browse that weeks playlist at my leisure, using the "star" feature in Navidrome to decide what to keep. Once I'm done I run another script manually that takes all of the starred tracks from those two playlists and moves them into my catalog, and then deletes the remaining tracks and the playlists.
This means I just need to go through and listen to recommendations and click a button on what to keep, and the rest is discarded automatically. It really doesn't get any simpler than this!
What remains will then be available for on-demand playback through Navidrome and also added to the full catalog that powers the 24/7 radio station.
FAQs from the last thread
What is being used to download from X? - spotdl is being used for Spotify.pytube is being used for LastFM and Youtube. spotdl is also just downloading tracks from Youtube under the hood.
What is the audio quality of the downloaded tracks? - Since these are coming from Youtube, everything is a 128kbps VBR Opus codec. It's certainly not FLAC but it's good enough for my enjoyment.
This is awesome. I’m gonna save this and unfortunately probably do nothing with it. It’s something I’d like to do but I don’t listen to a whole lot of music and spotify is cheap enough it’s not worth the effort for me. But this is quite impressive.
It's good enough for my purposes, which mostly involve streaming it over 96kbps for playback on wireless headphones. It's a small price to pay for the convenience of the automation.
Spotify pretty much has them down from my years of use. Even if you're not coming back and listening regularly it will still update that playlist every week.
LastFM is getting my actual up to date listening habits as I use their scrobbling service with my music clients, including Navidrome.
Since these are coming from Youtube, nearly everything is a 320kbps MP3.
Where are you getting that format and quality from? I use youtube-dl and non-premium audio for everything is ~128kbps VBR Opus, which aligns with other sources I can find. If I use Premium then I can get up to ~256kbps m4a but still not 320 and I see no mention of 320 availability online. Is there a trick you are using?
That was my mistake, and I'll edit the post. I just verified that you are correct by checking a random subset of the MP3's I have. I clearly got wrong information from somewhere.
Interesting! I've been thinking about doing something similar with Youtube videos, make a little fake TV channel. I'll definitely be taking a look through your code.