Hey everyone, working on TuneLog (a ListenBrainz CF playlist generator for Navidrome) and looking for some advice.

Right now, ListenBrainz suggests 1,000 songs via CF, but only about 300(out of 9k songs) of them actually match the files in my local library. That’s fine for casual listening, but heavy listeners run out of fresh tracks too quickly.

I thought about pulling the CF recommendations of similar users to expand the pool. However, since CF inherently already uses similar users to generate my initial 1,000 tracks, I suspect doing this will just result in massive overlap and be a waste of API calls.

Should I bother adding this feature, or is the math too redundant? How do you guys handle keeping recommendations fresh when the local library doesn't have enough of the suggested tracks?

-- post body written by gemini ai after i gave it a proto type

I use navidrome and a script to auto generate the playlist, but every time it generates I have to manually sync the playlist, is there a way to automate this? The script doesn't create a new playlist it updates the existing playlist

To get sycned lyrics from spotify for the songs that have lyrics

1. open spotify
2. Go to the Inspect by clicking F12
3. go to network tab
4. in the search/filter type color
5. when the song loads the lyrics it will show up in network tab as a request(api)
6. select the response
7. go to reponse tab
8. expand the lyrics section in json, confirm its that songs lyrics
9. copy the object and make use of AI,
10. Prompt ai to "Give me synced lyrics format of this, make it in markdown code"
11. open your metadata editor/lyrics editor, (LRGET) select the song, click write, go to sycned tab and paste it
12. save, publish it if you wanna help the commuinity and done

https://preview.redd.it/dug8z5iay50h1.png?width=1913&format=png&auto=webp&s=24cad886a8e26bbff9b9439df835d9af9014d5be

To get sycned lyrics from spotify for the songs that have lyrics

1. open spotify

2. Go to the Inspect by clicking F12

3. go to network tab

4. in the search/filter type color

5. when the song loads the lyrics it will show up in network tab as a request(api)

6. select the response

7. go to reponse tab

8. expand the lyrics section in json, confirm its that songs lyrics

9. copy the object and make use of AI,

10. Prompt ai to "Give me synced lyrics format of this, make it in markdown code"

11. open your metadata editor/lyrics editor, (LRGET) select the song, click write, go to sycned tab and paste it

12. save, publish it if you wanna help the commuinity and done

https://preview.redd.it/32oym8eyx50h1.png?width=1913&format=png&auto=webp&s=a0ed132fac9d3786281f865aa3baf5fb44419f85

I am using tailscale to access navidrome outside of lan, but I have to switch the server everytime, is there any client on Android that auto switch server url depending on connectivity?

Concept : collaborative filtering

The idea is like Spotify recommendation "User X & Y" approach: If User X and User Y have highly similar tastes, and User Y loves a specific track, User X will probably love it too.

To achieve this in a self-hosted environment, the idea leverages ListenBrainz's user compatibility scores:

1. Find Matches: Query ListenBrainz to find global users with a high compatibility score to your profile.

2. Extract Favorites: Pull the top or most-repeated tracks from their listening history.

3. Local Cross-Reference: Scan those tracks against your local Navidrome database.

4. Recommend: Take the tracks that actually exist on your server and inject them into your dynamic playlists.

pros:

• better recommendations

Cons:

• if the song is not in your local library then either download or skip

Is it worth building? Should I build it?

Concept

The idea is like Spotify recommendation "User X & Y" approach: If User X and User Y have highly similar tastes, and User Y loves a specific track, User X will probably love it too.

To achieve this in a self-hosted environment, the idea leverages ListenBrainz's user compatibility scores:

1. Find Matches: Query ListenBrainz to find global users with a high compatibility score to your profile.

2. Extract Favorites: Pull the top or most-repeated tracks from their listening history.

3. Local Cross-Reference: Scan those tracks against your local Navidrome database.

4. Recommend: Take the tracks that actually exist on your server and inject them into your dynamic playlists.

pros:

• better recommendations

Cons:

• if the song is not in your local library then either download or skip

Is it worth building? Should I build it?

Image for refrence i want it to split the artist for multiple artist, but why is it not doing this? i have added it in tags.artist.spliit

# Global Settings

Agents = "apple-music,deezer,lastfm"

[Scanner]
# Split rules for multiple artists
Tags.artist.Split = [", ", " & ", " / ", " feat. ", " ft. ", "; "]
Tags.albumartist.Split = [", ", " & ", " / ", " feat. ", " ft. ", "; "]
Tags.artist.Aliases = ["artist", "artists", "tpe1", "author"\]
Tags.albumartist.Aliases = ["albumartist", "album artist", "tpe2", "aart"]

[LastFM]
ApiKey = ----
Secret = ---
Enabled = true

My tunelog project handles the song rating by it self if a song has 3 listen history and they all are complete then the song gets a 5 stars automatically.

Should I do the same for album? If album songs has 5 star rating then album gets a 5 star rating?

Right now, I use LRCGET to get the lyrics but that's is not enough, it's massive flaw is messy metadata, it can't find lyrics if the metadata is diffent

I tried picard lyrics plugin but that seems to not work for me?

Is there anything else I can use to automate this?

Right now, I use LRCGET to get the lyrics but that's is not enough, it's massive flaw is messy metadata, it can't find lyrics if the metadata is diffent

I tried picard lyrics plugin but that seems to not work for me?

Is there anything else I can use to automate this?