Music metadata tagging explained: the tags that matter, Vorbis comments vs ID3, and how MusicBrainz Picard makes your library actually searchable.
You know an album is in your library. You've played it. But you type the name into search and get nothing, because it's filed under Sufjan Stephens, or the album field says Unknown Album, or half the tracks think they're by Various Artists and the other half don't. The files are fine; the labels on them are a mess.
That's the quiet truth about music metadata tagging: it's not a chore bolted onto collecting, it is the collecting. The audio is just samples. Everything your player knows (who made this, what album it belongs to, what order the tracks go in, what the cover looks like) lives in a few kilobytes of text riding inside each file. Get those right and your library feels like a well-run record shop. Get them wrong and you own a very large pile.
There are dozens of possible fields. Eight of them do nearly all the work:
feat. vs ft. vs featuring.Various Artists while each track keeps its real performer. Most "why is my library shattered into single-track albums?" mysteries are a missing album artist tag.Everything else (composer, label, catalog number, lyrics) is optional depth. Nice to have, never required, and better left empty than filled inconsistently.
Here's where formats matter, because your FLACs and your MP3s store all of this differently, and the difference explains a lot of weird player behavior.
FLAC uses Vorbis comments. These are charmingly simple: plain UTF-8 text, FIELD=value, one per line inside the file. ARTIST=Portishead. ALBUMARTIST=Portishead. Any field name is allowed, you can repeat a field for multiple values (two ARTIST entries for a duet), and there are no encoding gotchas because everything is UTF-8 by definition. It's the tagging system you'd design today from scratch.
MP3 uses ID3v2. This is an older, binary system where each piece of metadata lives in a frame with a four-character ID: TPE1 is the artist, TALB the album, TPE2 the (de facto) album artist, APIC the embedded picture. It works (billions of files prove it) but it carries history: two incompatible-ish versions in the wild (ID3v2.3 and v2.4), multiple text encodings, and players that disagree on the fine print. Many MP3s also carry a vestigial ID3v1 tag: 128 bytes bolted onto the end of the file, with fields brutally truncated at 30 characters. If you've ever seen an album title cut off mid-word on an old device, you've met ID3v1.
ALAC and AAC files in M4A containers use yet a third system (MP4 metadata atoms), and WAV barely supports tagging at all, one of the practical reasons collectors prefer FLAC for archiving.
The saving grace: good tagging software hides all of this. You edit "Album Artist" and the tool writes ALBUMARTIST= in a FLAC and TPE2 in an MP3. But when a tag mysteriously shows in one player and not another (especially album artist on MP3s) the format seam is almost always the culprit.
You should not be typing this stuff by hand. MusicBrainz Picard is the standard tool, it's free and open source, and it changes tagging from data entry into review.
The workflow: drop your files in, cluster them into albums, and hit lookup. Picard matches them against MusicBrainz (a huge, community-maintained database of releases) and proposes complete, consistent tags: exact titles, correct years, proper track ordering, album artist handled correctly, cover art fetched automatically. For mystery files with no tags at all, AcoustID fingerprinting identifies tracks from the audio itself, which feels like magic the first time it names a file called Track07.mp3.
Two honest caveats. The database is community-built, so a very obscure release may be missing or imperfect; you review before saving, always. And Picard will happily rewrite thousands of files, so agree with its choices before pointing it at the whole library. Better yet, tag at the moment music enters your collection: ripping CDs with a MusicBrainz-aware ripper means files are born correctly tagged and never need rescue.
Tag once, at import, from a database, and review it. That single habit replaces every future weekend of cleanup.
Perfect individual tags still make a bad library if they don't agree with each other. The failure mode is drift: R.E.M. here, REM there; dates as 1994 on one album and 1994-10-04 on the next; one compilation credited to Various Artists, another to VA. Each is fine alone; together they shred browsing and search.
So decide your conventions once (date format, genre vocabulary, how featured artists are written) and let the tools enforce them. This is also where tags and folder structure meet: treat tags as the single source of truth and generate filenames from them, so a fix in one place fixes everything.
One more field pulls serious weight: embedded cover art, an APIC frame in MP3s, a PICTURE block in FLACs. Art embedded in the file travels with it everywhere, which folder images can't promise. Sizing, sourcing, and the embedded-versus-folder question deserve their own discussion; there's a full guide to album art for that.
Here's the payoff for all this care: a well-tagged library needs nothing else to work. Drag a folder into the player and titles, artists, albums, and artwork appear instantly, read straight out of the files, no database, no account, no cloud. The metadata is the interface.
Tags are a letter to your future self. Ten years and three computers from now, the files will still open; the only question is whether they'll tell you what they are. Spend the kilobytes.
FLACCID plays your local FLAC, ALAC, WAV and more right in the browser: no upload, no account, no streaming.
Launch the Player