Trying out Picard, but need improved matching!


I have ripped my CDs to FLAC (using dBpowerAmp), and does thus have almost 14000 tagged FLAC-tracks on my hard drive (~400 GB).

I thought it might be nice to add tags and connect them to the corresponding release and artist in Musicbrainz, so I downloaded Picard, selected all tracks, and asked it to look them up for me.

Results were unfurtunately underwhelming so far. After around 9 hours of work, despite the tracks being organized in a directory per album, and tagged with correct artist and title, over 3500 tracks are not matched, and it tries to change other tracks from being connected to the correct album to being “non-album tracks”, even though I can find track on my album in your online database.

So far I’m thus quite disappointed with the results, but perhaps you know of some trick that I should try?

If you want me to send logfiles for you to improve Picard I would be happy to do so, but right now using it is not an option for me.

Hi @endian!

Firstly, the classic Picard/MB disclaimer…

Picard isn’t the mythical one-click wonder. Work in batches and check your matches.
Unless your tags and our database are both perfect, Picard wont get everything right. If you want exact results (eg care about which edition of a release you’re tagging something as), do some manual checking. If you don’t care about stuff like that, you’re probably fine to just mass-save, but I wouldn’t.

Sorry about that :stuck_out_tongue:
Moving on…

First question:
How are you matching tracks?
Using ‘Lookup’ will use existing tags to try to find a match in the database.
Using ‘Scan’ will ignore existing tags and just search using the audio fingerprint (so some of your results are expected if this is how you’ve been matching stuff)
As far as I know none of these take into account directories/folder structure.

Second question:
Are you tagging obscure/niche music?
Not all music is on MusicBrainz, and a lot of it hasn’t had any fingerprints submitted (if you’re using ‘scan’).
If everybody adds stuff that’s missing this wont be a problem in the future, time and enthusiasm permitting ; )

In any case, a simple tip is to change these settings in Picard > Options > Metadata > Preferred Releases:

1 Like

Thanks for your reply, Simon.

This was what I did:

  1. Took a copy of my FLAC tracks to a local hard drive.
  2. Start Picard, and import the tracks, (Takes a while, but not hours)
  3. Select all tracks.
  4. Press the Lookup button.
  5. Wait 9 hours or so for results.
  6. Voila!

I’ll try to change the settings you mention, tonight, and let you knwo what happens!

The mismatching CDs were not very obscure, they included Abba, ABC, Adele, etc. They were correct in your database on the web (at least the ones I checked).

/ Kenneth

Great, if they’re in the DB and your tags are ok, hopefully things will be pretty straightforward!

Maybe do just some folders while you’re trying to get different settings to work :anguished:
Like some of the popular ones that didn’t match - perhaps there’s some quirk in the existing tags, eg different spelling of album titles etc?

By the way, if you hit ‘save’ on that many files at once, Picard/your computer might freeze up, and then you’ll have to start over. Anyway, even if everything looks matched/gold, I would still be very very cautious about doing this with that many files.

A good workflow can be to try ‘lookup’, and then try ‘scan’ on anything that didn’t get matched.
You’ll end up with a lot of various artist albums etc, but as long as one track from each album is matched correctly, you can drag and drop everything into the right place fairly easily.
Working with hundreds or thousands of albums at once makes it pretty hard to untangle though, you’ll have way too many albums in the right hand pane to figure out what goes where. Perhaps move your saved/ ‘easy’ albums into another folder and then you can later spend some time on just the tricky ones.

The maybe most important tip is to use Cluster before doing Lookup. This will sort the loaded files by album and lets Picard lookup an album as a whole instead if the individual tracks separately.

Also work in batches if a couple if albums and do not load your entire collection at once.

There are a couple of beginner threads already explaining this, search the forums :slight_smile:

1 Like

Oops, if you’re not clustering, that’s definitely causing things to get split up!
Didn’t think of that at all… I do it too automatically :open_mouth:

I tried your original idea tonight, and results were perhaps a little better, but not great.

For example on a-ha Hunting High and Low seven tracks were matched while three were left unmatched. Adele 21 had two unmatched tracks.

In my view, the original matching should have been able to find those too, but I’ll have to try to see if this clustering helps, though.

1 Like

Hi again,

I tried the clustering today, and that helped!

My question is just why it’s not made automatically?

/ Kenneth