False auto tagging (Look-Up or Scan)

I have a database with several thousand songs and I want MusicBrainz Picard to auto tag my MP3 files and find the right cover art. In many cases Picard auto tags completely wrong. Picard even has a tendency to tag songs with “Best of…” albums or collection album instead of the original album. And even when I change the meta data manually Picard will not find the right cover art. I have tried this with several of the best know albums and songs of all time, e.g. “Beat it” from Michael Jackson. Picard tags this as a standalone recording. Then I changed it manually, added the album “Thriller”, album-artist, release data, track number, etc. When I click on on “Lookup” or “Scan” Picard declares this still as a “standalone recording” even after I deleted all the MusicBrainz IDs. How can this be solved?

1 Like

Welcome to the Picard community.

Have you read the “how to” documentation at e.g. When files are grouped by album — MusicBrainz Picard v2.8.3 documentation ?


The process that Sophist linked is indeed the ideal one imo.

One thing that is not made immediately clear by Picard is that the ‘scan’ button ignores all existing tags and ‘listens’ to the song to find a match. This means it will return a lot of compilations etc. I would only use it as a last resort. If you are using it, you can find settings in the options that change what it will return - e.g. drag the compilation slider all the way to the left.

In the options you will also find a setting to ignore existing MBIDs.

Finally: if you’re not already, please proceed slowly and even start with a backup of a selection of files. Changes can’t be undone!

Go into Picard OPTIONS \ METADATA \ Preferred Releases and mess with the sliders.

Give the Singles \ Albums \ EP a higher priority in the match. And drop the Compilations down lower. This adjusts the bias to the matches.

1 Like

Thanks for your help. I does not work. I even added several other Michael Jackson songs from the same album, provided meta data like release year, track number and even the correct cover art manually with another editor, saved all the data. So I was at the point that I did not even need Picard anymore because I had already done everything manually. Ithen launched Picard. Picard wanted to replace a lot the correct information with incorrect meta data and incorrect cover art. It’s scary to see that Picard, instead of enhancing the data, even fucks up a lot of meta data that are correct.

“Lookup” needs tags like artist name, track name, album name, maybe track number.

Are you using the Cluster button first? Putting files together into their albums again helps the algorithms.

But Picard is not really a tool for bulk correcting tags on a heap of random files from multiple sources. It works best with albums and clean rips. Whatever the source, it will always need manual checking as no algorithm can get a perfect hit rate. I don’t know what the stats are, but I doubt it is better than 90% and will drop with lower quality files.

What state are your files currently in? For example, if you have filenames and folders setup but no tags, look at MP3TAG and use the Filename to Tags buttons to pre-fill your tags with something searchable. It may also be worth looking at MP3TAG to grab the artwork too.

Picard is designed to be fussy and find exact album matches aligned with how Releases are organised in the MusicBrainz database. But if you just want any good art and clean tags, MP3TAG may be easier to work with.

If you’re using the “Scan” button, that looks up the metadata based on the audio and ignores any tags in the files. You shouldn’t need to use that unless your music files have incorrect or no tags and you honestly have no idea what you’re listening to.


Such comment makes me think the problem might be elsewhere.

Do you actually Cluster your files (as recommended in documentation linked above)?
Did you change any option? which ones?

Can you explain clearly what’s your workflow?
Please provide screenshots of each step, a debug log might help too (see documentation on how to get such log).