2. Even when you do import a full album, it often struggles to match tracks. Here's part of a transcript I saved:
Match (29.3%):
StarSystems - Lift
≠ artist, album, tracks, unmatched tracks
MusicBrainz, CD-R, 2019, GB, None, None, None
https://musicbrainz.org/release/7c68b7ff-27c0-4da7-8430-d4228c6f6ff8
≠ Artist: -> StarSystems
≠ (#0) StarSystems - Arrival [2238026155].mp3 (1:08) -> (#1) Arrival (1:08)
≠ (#0) StarSystems - Slipstream [1268364723].mp3 (6:23) -> (#2) Flower (6:25)
≠ (#0) StarSystems - Trigger Fish [1805252711].mp3 (5:06) -> (#3) Party Time (5:03)
≠ (#0) StarSystems - Moth [857985311].mp3 (4:51) -> (#4) Moth (4:51)
≠ (#0) StarSystems - Vortex [1717930774].mp3 (5:38) -> (#5) Vortex (5:38)
≠ (#0) StarSystems - Flower [3129604796].mp3 (6:25) -> (#6) Slipstream (6:23)
≠ (#0) StarSystems - Party Time [545976708].mp3 (5:03) -> (#7) Trigger Fish (5:06)
≠ (#0) StarSystems - Lift (feat. Jordan Rudess) [3330158657].mp3 (8:09) -> (#8) Lift (feat. Jordan Rudess) (8:09)
≠ (#0) StarSystems - Disengage [526901544].mp3 (2:01) -> (#9) Disengage (2:01)
Unmatched tracks (1):
! (#0) (5:06)
Apply, More candidates, Skip, Use as-is, as Tracks, Group albums,
Rescan directory, Enter search, enter Id, aBort?
Look at tracks #2, #3, #6, #7. Why is it trying to switch the names around, even though the original name + length is right, and the new name + length is wrong?
Picard has worked flawlessly for me so far, it's just a shame it can't be automated.
I've used it to successfully organize my messy library of (then) 50 thousands songs back in 2014. It's great if you can wrap your head around the weird ass workflow.
The plugins/extensions have also been proving quite useful, there are now quite a lot more available and useful ones. I've managed to incorporate some into my workflow.
I've got a large, local library in Apple Music—the application, not the service, they need a naming czar—amassed over decades of purchases. The auto-lookup from iTunes/Music is fine, but I find a real joy in having correct metadata. All those mix CDs from high school!
I care less about high-quality album art than Apple execs do, I think, and I likewise don't bother too much with all the meta tags, but it's surprising how often the CDDB or Gracenotes or whatever Music.app uses are wrong, have typos, miss information, etc. I scrobble listens, so it's a natural progression for anything new or clearly wrong.
I think of it like tending a garden, and find it quite peaceful. It's clearly not for Regular People™ like so much else here. Every other person in my life is happy with "Unknown Artist - Unknown Album - Track 2" or just using a streaming service.
I am. As the other commenter mentioned the workflow is very odd, and the UX of the app could do with massive improvements (haven't checked V3 yet), but it does a good job and I'm grateful to the devs for it.
I just recently went through my whole library with Picard and it worked well while the UI was definitely geared for power users. Really happy with the end result.
There was one UX papercut that ended up messing some data on tracks that I had to correct later: fields with multiple values, like "Artists", show the multiple values like "Beats and Styles; Toni W.". But the field in the UI is also editable as a text field, and if you manually write that with the semicolon, it stores one artist with the semicolon included! It took me a while to realise that I can right click to open an editor where multiple values can be entered and edited separately.
Picard was extremely useful to me when I decided to put all my audio on a streaming server in a local container (Navidrome). I used it extensively about five months ago. I find MP3 tags (and all the various formats and syntax) a nightmare usually but this was bearable. Picard was great.
I love MusicBrainz for everything they did but I don't think using a tagger as a separate program you have to open and point to a file is good UX, even if the tag lookup is fully automated.
It should be a daemon that auto tags all new files in the background and only prompts you when there's a problem. And if it finds a new song with metadata that's not in the database you get the option to upload it. Even better if it scrapes CDs and streaming services. Maybe I'll build it one day.
Ehhhh, people are specific about their albums. I recently used this auto tagger and it would tag a song in a strange album. Like song Ocean Avenue from Yellowcard was not from their album Yellowcard but some compilation album that I had never heard of.
Tons of people have scripted Beets to autotag music as they're downloaded or appear in a directory though. So in that sense it's an automatic, non-manual workflow
I'm personally a fan of the tagging being a separate explicit action, it suits my flows at least: Download to staging folder -> Picard (tags, grabs artwork, normalises file names, I intervene where needed) -> automatic transfer to streaming server when it's finished.
Man, I could be wrong, but I would venture to guess that the number of people using this has decreased massively. In my teens and 20s I remember religiously scanning my entire MP3 collection through Picard and also gleefully scanning disc IDs into the database. Haven't used MusicBrainz in over ten years at this point.
Anecdotally - I, and several people I know, have independently started using CDs, Soulseek and Bandcamp to rebuild personal audio collections after becoming annoyed with streaming apps. LLMs have really helped organizing and tagging music efficiently.
I got back into it around a year ago, highly recommend. Very refreshing to experience parts of the old world that have survived like soulseek and torrents again.
Yeah I'm switching from Spotify to building my library manually now. I like having the full control, personal choice of music app/client for each platform I listen on, the ability to individually adjust tracks to my personal liking (some songs need manual ReplayGain adjustments to sound correct), being able to have my favorite albums be in FLAC and the rest be good quality Vorbis... it's all just better, except for discoverability. I've found so much good music through Spotify's algorithms that I probably never would have found otherwise. I wish a good FOSS alternative for that existed.
I am doing the same. I started to connect my agent to beets db. Actually would be quite I interested in some clean-up skills. I actually started to go to the local library again to add to my music collection (at least in my jurisdiction you can do private copies of non DRM content)
Could you provide more details about your LLM + Beets setup? I've been thinking of doing something similar and figured Beets could help me organize the random piles of MP3s I have scattered around my drives, but I don't want to manually correct the ambiguities from incorrect tagging.
I just started around messing with opencode and beets cli. No MCP, no real skills etc.. Worked reasonably well on some retagging, better dir layout. Nothing special. Probably could also done it without an agent. I still fail to clean up some ripped apart compilations, etc. (Partially due to premature re-sorting). Trying to figure out the human oversight part. My goal was also doing a sensible genre and other semantic description over the whole collection to enable RAG style playlists, but I guess one also has to have a decent plan for such things. I was actually hoping there was already more out there.
Multiple artists, genres, moods all behave different for each player. I hate thinking about tagging music.
Can't say I've had a good experience with picard either. Too many versions for many songs means tags can be anything if you leave it alone and do its thing
Recently me and a bunch of DJs reinvented how we manage metadata for our collections.
The TL;DR is use SQLite for everything. Both a local central database describing the entire catalog, and one “sidecar” database for every song representing its individual metadata.
The huge advantage is it stores metadata that ID3 tags never imagined like beat grids and phrases and cue points.
The other advantage is agents crush manipulating data that is in SQLite. Chatting with my music collection feels better than any music library app did.
Of course tools also import and export to native audio file tags and use the musicbrainz API.
Looking for feedback and other music nerds that want to collaborate on this approach.
This is interesting, thanks for sharing. It looks like all the hamburger menu links are leading to a 404 for me on mobile though (didn’t try desktop).
Does the “analysis” tab in this app imply that this software does its own cdj compliant beatgrid and phrase/key analysis? If so, do you find it works better than Rekordbox?
It's unfortunate that not all of their official plugins were ported before the update was released. I do use a few that aren't available (AcousticBrainz Tags, Wikidata Genre, etc.) and that makes it difficult to choose the update and then missing out on the tags.
EDIT: I installed it on a separate machine, and there is a setting in the wizard called "Fetch detailed track relationships". I'm unsure if that means it will contain AB tags or the Wikidata genres though.
Welp, looks like it's time to re-rip / re-tag my CD collection at higher quality. Gives me a good reason to dust off some stuff (physically and virtually) that I haven't listened to in a while.
Congrats to Philipp and the guys on the release, and thanks for keeping it alive. I was just looking at the MusicBrainz blog today, went down the history lane. I find it unbelievable that it's 20 years since I took over the project from Rob and did my first release with him.
I was in the Uni in grad school. That one year, the university decided to wire up all dorm rooms with ethernet. And guess what? Everyone's SMB server was wide open.... and in the lab I had access to desktops with massive hard drives, hooked to the university's ethernet. A few scripts using SAMBA was all that was needed to suck up the entire dorm's collection of songs, videos, porn, etc. , running into many GB of data. Then I used Picard to tag the audio files and rearrange them into directories by genre, artist, etc.
Good stuff! I'm excited to check it out. I see a lot of other comments here wondering if others still use this... I like MusicBrainz Picard because it makes it really easy to see everything that's going on at once. View the tags, see changes that will be applied, easily switch between configuration profiles, and easily browse to the recording or release page in MusicBrainz.
The manual workflow might not be for everyone, but it's a fantastic GUI application.
This is awesome. I've been using Picard for years and it's great. I just went through my entire library to add synced lyrics (via plugin) and fix my naming script. The new features for synced lyrics will help a lot in the future.
trencedamp | 6 hours ago
ZetaRicky | 5 hours ago
sdumi | 5 hours ago
jamesbvaughan | 5 hours ago
cdrnsf | 5 hours ago
nvme0n1p1 | 5 hours ago
1. Tagging doesn't work if you don't import every song in an album at once (see https://www.reddit.com/r/musichoarder/comments/1j0wnbz/need_... and dev comments on https://github.com/beetbox/beets/issues/5014)
2. Even when you do import a full album, it often struggles to match tracks. Here's part of a transcript I saved:
Look at tracks #2, #3, #6, #7. Why is it trying to switch the names around, even though the original name + length is right, and the new name + length is wrong?Picard has worked flawlessly for me so far, it's just a shame it can't be automated.
cdrnsf | 3 hours ago
amlib | 5 hours ago
The plugins/extensions have also been proving quite useful, there are now quite a lot more available and useful ones. I've managed to incorporate some into my workflow.
Amorymeltzer | 5 hours ago
I've got a large, local library in Apple Music—the application, not the service, they need a naming czar—amassed over decades of purchases. The auto-lookup from iTunes/Music is fine, but I find a real joy in having correct metadata. All those mix CDs from high school!
I care less about high-quality album art than Apple execs do, I think, and I likewise don't bother too much with all the meta tags, but it's surprising how often the CDDB or Gracenotes or whatever Music.app uses are wrong, have typos, miss information, etc. I scrobble listens, so it's a natural progression for anything new or clearly wrong.
I think of it like tending a garden, and find it quite peaceful. It's clearly not for Regular People™ like so much else here. Every other person in my life is happy with "Unknown Artist - Unknown Album - Track 2" or just using a streaming service.
have_faith | 5 hours ago
Ndymium | 5 hours ago
There was one UX papercut that ended up messing some data on tracks that I had to correct later: fields with multiple values, like "Artists", show the multiple values like "Beats and Styles; Toni W.". But the field in the UI is also editable as a text field, and if you manually write that with the semicolon, it stores one artist with the semicolon included! It took me a while to realise that I can right click to open an editor where multiple values can be entered and edited separately.
sherr | 5 hours ago
brlcad | 5 hours ago
tancop | 5 hours ago
It should be a daemon that auto tags all new files in the background and only prompts you when there's a problem. And if it finds a new song with metadata that's not in the database you get the option to upload it. Even better if it scrapes CDs and streaming services. Maybe I'll build it one day.
HaloZero | 5 hours ago
disposition2 | 5 hours ago
Someone beet you to it [1]
1. https://beets.io/
nvme0n1p1 | 5 hours ago
krisbalintona | 4 hours ago
cdrnsf | 4 hours ago
have_faith | 5 hours ago
skarz | 5 hours ago
dSebastien | 5 hours ago
sp1nningaway | 5 hours ago
arvid-lind | 4 hours ago
seanw444 | 4 hours ago
riedel | 3 hours ago
RussianCow | 3 hours ago
riedel | 32 minutes ago
thekid314 | 5 hours ago
Time to pull up Bandcamp and Soulseek and keep adding to my collection.
pluc | 2 hours ago
ThatMedicIsASpy | 2 hours ago
Multiple artists, genres, moods all behave different for each player. I hate thinking about tagging music.
Can't say I've had a good experience with picard either. Too many versions for many songs means tags can be anything if you leave it alone and do its thing
nzoschke | 5 hours ago
Recently me and a bunch of DJs reinvented how we manage metadata for our collections.
The TL;DR is use SQLite for everything. Both a local central database describing the entire catalog, and one “sidecar” database for every song representing its individual metadata.
The huge advantage is it stores metadata that ID3 tags never imagined like beat grids and phrases and cue points.
The other advantage is agents crush manipulating data that is in SQLite. Chatting with my music collection feels better than any music library app did.
Of course tools also import and export to native audio file tags and use the musicbrainz API.
Looking for feedback and other music nerds that want to collaborate on this approach.
https://deadca7.com/blog/deadca7-open-music-database-specifi...
carljungslabtek | 2 hours ago
Does the “analysis” tab in this app imply that this software does its own cdj compliant beatgrid and phrase/key analysis? If so, do you find it works better than Rekordbox?
asymmetric | an hour ago
And what are the chances that one of the big names (primarily Pioneer) take up something like that?
liotier | 5 hours ago
rmnclmnt | 3 hours ago
PenguinRevolver | 5 hours ago
EDIT: I installed it on a separate machine, and there is a setting in the wizard called "Fetch detailed track relationships". I'm unsure if that means it will contain AB tags or the Wikidata genres though.
cbm-vic-20 | 4 hours ago
lukaslalinsky | 3 hours ago
infofarmer | 3 hours ago
it was good to work with you guys back in 2011-2012
adfm | 2 hours ago
mlmonkey | 2 hours ago
I was in the Uni in grad school. That one year, the university decided to wire up all dorm rooms with ethernet. And guess what? Everyone's SMB server was wide open.... and in the lab I had access to desktops with massive hard drives, hooked to the university's ethernet. A few scripts using SAMBA was all that was needed to suck up the entire dorm's collection of songs, videos, porn, etc. , running into many GB of data. Then I used Picard to tag the audio files and rearrange them into directories by genre, artist, etc.
retrodaredevil | 2 hours ago
The manual workflow might not be for everyone, but it's a fantastic GUI application.
curtisblaine | an hour ago
koenada | an hour ago