Picard 2.x and genres. Good practices? Advice on plugins?

Wow, that was fast! I am trying it out and so far it seems to work as expected. The " / " separators appear correctly. I had to restart Picard to get the new plug-in to install and to see its Options, but I suppose that’s about the app not the plug-in itself.

1 Like

Hi

The genre base grows nicely and thus some problems appear, imho. We have several levels of genres, they’re in:

  1. Artist
  2. Release group
  3. Release
  4. Track

For example, let’s take an album:

Picard:
obraz

The problem is that: 4 of 13 tracks has different genres. I personally prefer Release or Artist tags. And thats why we need in Picard something like in Last.fm plugin:

obraz

but I would expand these options additionally:

  • Artist tags
  • Release group tags (when the Release does not have tags it will use it from Release group)
  • Release tags
  • Tracks tags

What do you say?

P.S.I forgot to add that I use only built-in Picard genres.

1 Like

I agree that e.g. “Fall back on album’s artist’s genres…” would benefit from having a separate tickbox for each type of relevant entity (Artist, release group, release, track… and maybe even label?) I’ve sometimes independently tagged the genre of specific tracks on an album, but when it comes to the more practical level of actually listening to the music, I’d like to keep each album’s tracks together, so I’d love to be able to untick this for tracks only.

1 Like

Indeed. Thankfully, Zas has implemented this (for MB genres, not last.fm ones), so hopefully that will be in a future version of Picard soon.

Very interesting work from @Zas, I only hope MB’s own database tag fields will be populated by more people real soon now …

Over the past so many years, I’ve been using fdemmer’s Last.fm.ng plugin quite successfully (with my own config.ini of course). Note this is not the same as Picard’s Last.fm plugin!

It already lets you specify a whitelist, categories, number of tags, tag translations, album genre fallback and whatnot and takes its data from last.fm. You can find my tagger script using it and the corresponding config.ini for the plugin in Repository for neat file name string patterns and tagger script snippets. I think it’s well worth a look and a deeper dive into its documentation.

A word of warning, though: Results can be horribly bad if there are several artists of the same name, like, say a pop group and a death metal band. You’d have to correct the genre and grouping tags manually in this case.

The plugin isn’t perfect, but it’s probably a good start until we can get more from the MB database. Be aware that it will slow down lookups considerably because of last.fm query rate limiting.

1 Like

Could you maybe reupload this? Would love to try, I’m kinda desperate here as nothing can find genres automatically for me.

I also can’t get last.fm.ng to run on a mac.

The latest version of the Wikidata plugin is available in the official Picard plugins repository. You can install it from inside Picard in Options > Plugins.

Best would be probably to ask the author of this plugin, @fdemmer, for help. Not sure if he is active here on the forums, maybe ask on GitHub - fdemmer/Picard-Last.fm.ng-Plugin: A last.fm-tag plugin for MusicBrainz Picard . But you can also open a new discussion here and provide details what exactly is not working for you, maybe somebody can help.

Please note that this particular plugin might be a bit more difficult to setup since configuration is done via a separate configuration file.

Anyone have any further advice on Genre’s? I’ve enabled the default last.fm plugin, and have experimented with the percentage, even pumped it up to 100%. However, I still get genres like “Fuck Yes” and “Better than Selena Gomez” and “Best Song Ever” from last.fm. It’s really aggravating. I understand there’s a blacklist, but with so many user-entered tags in last.fm there is no way I can catch them all.

For example, this very well known song: https://musicbrainz.org/recording/e0afc851-6df3-42b9-877e-4575200a23c5 pulls the Genre from last.fm as “Bhangra / Epic / Masterpiece / Fuck Yes” with minimal tag usage set to 100%.

I tried Folksonomy, but the tags are pretty inconsistent and the hit ratio is super low compared to last.fm. I kept getting “Jazz” as a genre for a funk album called “100% Funk” that is all funk music. I suppose it might technically be correct in a way to say it’s Jazz, but it definitely should at least include “Funk”.

I switched to wikidata-genre too, but as a poster in here already said it takes a ton of API calls and forever to get a result. When I tried scanning 25 songs with wikidata-genre enabled, it went up to 400 requests. With last.fm instead of wikidata-genre, it was only at 50 something requests. That’s an insane number of requests to get a basic piece of data (but I don’t doubt it has to do with the architecture of the API, and there’s nothing the plugin can really do).

I may have to keep using wikidata-genre to bulk scan files and then return to my computer an hour later to actually commit the new data once it’s all pulled.

EDIT: I am actually on version 2.0.4 released about a year ago – I am going to upgrade and see if that makes a difference!

EDIT: The update didn’t really help at all for Genre’s unfortunately.

EDIT: Jim Henson’s Fraggle Rockin’ CD is tagged as “Songs About Masturbation” by last.fm … I don’t think I can use last.fm anymore.

1 Like

My last comment hopefully for a while – I am really liking the genre’s straight from MusicBrainz and am probably going to go that route, editing in any genre’s that are missing as I go.

4 Likes