Expanding Roon metadata and tags

I’m starting to play with ChatGPT (and other AIs) to analyze my music library and create playlists, etc. I’ve exported artist, album and track info and see that info like record label is flat out missing.

I’m okay leveraging tags to accommodate some of these missing fields, but wonder if others have already come up with solution for some of these issue? Note, that I currently cannot run Roon extensions, so I’ll handle all of my data processing off platform.

Here are a couple examples of things I’m interested in exploring:

  1. Provide a list of SubPop albums (EP + LP) up to 1995 - Discogs API has these answers.
  2. List all SubPop artist that were picked up by a major record label. Indicate the last SubPop release as well as the following album and record label. (Discogs API)
  3. My thought is to simple tag each album with record labels?

Other examples:
Add BPM tag to every track
Expand/refine genres
Backfill missing City/Country info

Anyone else playing with these ideas? Thoughts?

I did this:

1 Like

I have a couple updates and a question:

How can I import a LibraryTracks.xlsx file back into Roon. Do I just find the file in the correct directory and replace it?

First, the successes:
I was able to export a Roon playlist for my entire library (~27k tracks) as well as the LibraryTracks.xlsx. I used both the .csv and .xlsx playlist files, which I’ll get to in a moment.

Given the size dataset, I broke the playlist file into 4 parts. Spotify only supports up to 11k tracks in 1 playlist and it’s their data I’m after. Soundiiz turned out to be a complete failure in that it only supports about 500 files per upload, but TuneMyMusic did a good job of getting most of my tracks into Spotify (90-95%).

To get the data out of Spotify playlist, I used a free site called Exportify.net. It was super smooth.

Associating the exported Spotify playlist with my original Roon playlist required ChatGPT (paid version). I did a bit of manual file cleanup before I ask the AI for help, but right off the bat, it was able to match about 15k tracks (song + artist). I was able to match another 5k or so using song + album. I spent days working with normalized data sets, “contains” logic and other fuzzy logic, but none of that turned out to be worthwhile. Given the size of the dataset, ChatGPT 4o crapped out more often than not.

It did do a good job helping me merge files, relabel and reprocess data, and finally… create a new LibraryTracks files with Spotify tags added.

But now that I have tons of fancy new data (tags) I don’t know how to get it back into the Roon platform. Thoughts?