Slashdot Mirror

MusicBrainz Picard is what you're looking for if you need to identify unknown audio files on your PC.

Unfortunately, a metadata "database" is only as good as the data you feed it. This may be limited by how much effort the end user is willing to put in.

Adding metadata to an existing music collection can be a huge PITA. For anyone tempted by the idea of doing it for their collection, I thoroughly recommend taking a look at MusicBrainz Picard. It's a cross-platform application that can analyse your music albums and attempt to match them against entries in MusicBrainz's vast database, populating your MP3s / FLACs with appropriate ID3 tags and cover art when it finds a match. If you have a large collection, it will still take some time and effort to get right, but you'll end up with something beautifully organised afterwards.

I still keep all my music (~20,000 tunes) in a simple folder / file structure, but every album and song is tagged correctly, along with cover art, so that I can also feed it to Clementine, Kodi, DNLA servers, Google Music, etc. Best of both worlds.

Unfortunately, a metadata "database" is only as good as the data you feed it. This may be limited by how much effort the end user is willing to put in.

Adding metadata to an existing music collection can be a huge PITA. For anyone tempted by the idea of doing it for their collection, I thoroughly recommend taking a look at MusicBrainz Picard. It's a cross-platform application that can analyse your music albums and attempt to match them against entries in MusicBrainz's vast database, populating your MP3s / FLACs with appropriate ID3 tags and cover art when it finds a match. If you have a large collection, it will still take some time and effort to get right, but you'll end up with something beautifully organised afterwards.

I still keep all my music (~20,000 tunes) in a simple folder / file structure, but every album and song is tagged correctly, along with cover art, so that I can also feed it to Clementine, Kodi, DNLA servers, Google Music, etc. Best of both worlds.

Having never heard of Discogs before, it basically sounds like it's like MusicBrainz with more data. Is that about right?

I'm always uneasy with helping private, for-profit entities fill a database with publicly-available information (e.g. Amazon-owned IMDb), since there's generally very little stopping them from taking it all private and locking it behind a paywall in the future. If a site is asking its users to assume the responsibility to generate and maintain the data, as is the case with these publicly-maintained databases, there should be protections in place ensuring that the data remains in the hands of the users. Wikipedia, MusicBrainz, and other non-profit projects do a great job of cataloging public information, and they do so with the backing of organizations and foundations whose primary purpose is to maintain the projects for their own sake, rather than to turn a profit.

I'm normally not an "information wants to be free" sort of guy, but apparently I am when it comes to this sort of information.

Forget Freedb. It's a mess of badly formatted and conflicting data. MusicBrainz is a much better alternative and has a cddb/freedb compatible gateway for software that doesn't support MusicBrainz directly.

http://musicbrainz.org/ is a good place to start, though Picard doesn't do all of the things you suggest.

I'd really like an auto-tagger and cover art finder that works in the terminal. Something like VortexBox that's not an OS, just a terminal command.
This is what I currently do (using long args for readability):

cdparanoia --verbose --batch --never-skip
flac --verify --best *.wav
rm *.wav

Then I open up MusicBrainz Picard and use this to get tags and cover art. Can this already be used in the terminal? There are tagging programs out there for the terminal, but they're either manual or don't auto-grab cover art, etc. It'd also be nice to specify a different source (musicbrainz for regular, something else for classical). Then I use SoundConverter to convert to OGG Vorbis or MP3, and it does a pretty good job of preserving tags. I just hate having to do so much GUI work. I'd rather just have one script file.

I gave up on torrents when I stumbled upon songs like Brown Eyes Girl by Jim Van Morrison or Red Red Wine by Neil Young.

That's why you use Musicbrainz Picard to boldly tag the files where no metadata has been before. Picard can usually recognise the songs (or better, albums) if they are at least somewhat popular.

In my case, I sit in front of the machine while ripping. I've got a lot of weird metal albums that aren't in Musicbrainz at all, have somewhat inaccurate information, or aren't titlecased properly (mostly poor titlecasing / ignoring the case used on the album where the case is actually significant). With sub five minute rips it's a quick process of pop the disk in, make sure the tag data is looking OK, and then pop the next disc in (well, at least once I finish beating abcde into submission and make the ripping part parallelizable like every other task is). And then comes the weird stuff, like multi-part songs that I want to tag this time around e.g. Gettysburg where each track should be marked PART="Gettysburg (1863)" (seemingly useless now, but with the data stored it'd be not-too-difficult to make something like xbmc pick the entire "movement" instead of just one part).

Since I'm going to be spending some time each week on this for months...

That's the point. "Album Artist", which is different from "Artist", is always the same in every track of the album, so grouping works.

For example, in the Pulp Fiction soundtrack, each music has a different Artist, but the Album Artist in all of them is "Various Artists", so the tracks are kept together.

Actually, Musicbrainz is quite open to this sort of thing. One of the suggestions I was given when trying to add a soundtrack which was included in a game's install (The Sims) was to put up a torrent and reference that, marking it as a bootleg. (I eventually convinced the naysayers that it was "official enough" as distributed by the game manufacturer)

Even more traditional albums released online are easy to get added (see: Santastic, The Kleptones, Wax Audio)

Actually, Musicbrainz is quite open to this sort of thing. One of the suggestions I was given when trying to add a soundtrack which was included in a game's install (The Sims) was to put up a torrent and reference that, marking it as a bootleg. (I eventually convinced the naysayers that it was "official enough" as distributed by the game manufacturer)

Even more traditional albums released online are easy to get added (see: Santastic, The Kleptones, Wax Audio)

Actually, Musicbrainz is quite open to this sort of thing. One of the suggestions I was given when trying to add a soundtrack which was included in a game's install (The Sims) was to put up a torrent and reference that, marking it as a bootleg. (I eventually convinced the naysayers that it was "official enough" as distributed by the game manufacturer)

Even more traditional albums released online are easy to get added (see: Santastic, The Kleptones, Wax Audio)

Last I checked, Musicbrainz wouldn't allow this sort of thing either. Mind you, specifically I was asking about bittorrent 'compiliations' of pre-existing material where, arguably, the set and ordering chosen results in a new work. I'm not sure if they would allow a torrent-only album under 'other' under the current practices:

            http://wiki.musicbrainz.org/ReleaseType [musicbrainz.org]

            But at least Musicbrainz is rather 'open' and allows dissent among the community on such topics--this leads to the obvious question, then: why isn't there a centralized 'open' metadata database like this for *all* forms of media: music, scores, movies, television, books, magazines, journal articles, encyclopedias, video games,etc...

http://musicbrainz.org/doc/PicardTagger

Sounds similar to what tools like MusicBrainz can do. http://musicbrainz.org/

No only Google: you can do it on your own PC!

The Picard Tagger from MusicBrainz can generate an audio fingerprint (PUID) from all files in a folder and then fetch the correspondent metadata from the community CC licensed music database.

How about a list of more apps?

• Calibre ebook manager
• Last.fm streaming music client
• VLC media player
• CDex CD ripping software
• MusicBrainz Picard for tagging audio files
• Pidgin IM client
• OpenPandora to put Pandora on your desktop and scrobble to Last.fm
• VirtualDub for simple video editing

Anyone else have any good recommendations?

http://musicbrainz.org/doc/PicardTagger This can be fairly successful, it's able to match tracks from audio finger prints as well as filenames and existing tags, typically successful to about 80-90% 100% tends to be mp3's already organised as albums. Still a major job to tag stuff, but a lot more efficient than doing it all by hand. No I don't have a perfectly tagged collection but its improved. unfortunately one area that will never be right is genre.

The MusicBrainz project could use them, and you get a tax write-off.

Why learn an arcane language like sh

Because sometimes it is exactly the right tool for the job. There are many tasks for which python or perl are overkill. Specifically, things which could be accomplished interactively on the command line (like repetitive tasks) but are better automated. Writing shell scripts also makes you more proficient on the command line, which is a valuable skill for any programmer or sysadmin. Your knowledge of command line tools and where/when to use them will surely increase.

#!/bin/sh

cat musicbrainz.xml | sed 's/>/>\n/g' | grep 'title>$' | sed 's/musicbrainz.txt

This is a quick and dirty script I use to convert a musicbrainz XML file (found on the details page of each album) to straight text. It's part of a collection of scripts which make up my custom tagging and music management system. It works great. But imagine what the equivalent script would look like in python or perl. A lot longer, for starters.

Editing metadata is not something someone is going to spend the hours doing, especially at the pay rate ($0).

Admittedly this probably won't help you personally, as it relates to released music, but Musicbrainz Picard (downloadable here http://musicbrainz.org/doc/PicardDownload) does a pretty amazing job of tagging music files properly for you. It uses some kind of audio fingerprinting tech so it's basically encoding independant.

I recently bought an iPhone (and a Macbook Pro) and was faced with the reality that I wouldn't be able to use Amarok with the phone as I did my Rockboxed iPod Nano (sigh) and that my music would need to be tagged properly for iTunes to behave. Picard handles it nicely.

Of course, it doesn't allow my iPhone to play vorbis files (that was my primary music format), or make iTunes any less of a slug. But that's a different matter entirely...

Assuming they're talking about something like MD5 hashing here--

Two rips of the same CD music track do not necessarily lead to the exact bit-by-bit identical MP3 file. Thus the hash is different, even if the same software, same CD, and same settings are used.

Two people with /exactly/ the same MP3 file will have the same hash. /Exactly/ the same, so if person B has added or changed ID3 tags, the file will already get a different hash.

There are other identification methods for music files, such as the one used by http://musicbrainz.org/, which /will/ provide the same hash for the same track even if it was ripped with different settings or on another computer. But from the article this is not what MediaSentry uses.

http://musicbrainz.org/ The data is either public domain or covered under Creative Commons. I believe the software is GPL. Definitely a better alternative to CDDB and freedb.

Musicbrainz...

Use MusicBrainz. All the cool kids are doing it!

Seriously. Musicbrainz was created after the CDDB fiasco (and FreeDB had its own share of problems). It operates under a non-profit organization to guarantee its freedom.
And on that feature bullet-point list, they add an API to recognize what that "Unknown Artist - Unknown Title.mp3" file you have.

"Crowdsourcing" usually means getting people to do stuff for you for free, where you own the results and the people who created them cannot use them except by paying you (if at all). This is why people should be sure that projects they contribute to as volunteers release their results under some kind of Free license. For example, contributions to Wikipedia are free-licensed, and even if Wikipedia died or turned ultra-evil tomorrow, you could use the articles yourself under the GFDL, or set up a fork based on them. The same is true of contributions to MusicBrainz (Open Audio License), among other such projects.

For a good early example of the opposite, recall the CDDB fiasco---lots of people submitting data that ends up owned by someone who won't let you use it except under onerous licensing terms. The rise of "Web 2.0" has basically taken CDDB-style business models and made them much more common, so it's important to make sure you aren't enabling that sort of thing that in the long term ends up working directly against your interests.

I used to be obsessed with tagging my MP3s. Then along came PicardQt, which finally makes MusicBrainz usable. The macros you use to specify the renaming pattern are still poorly documented, but otherwise it's superb. Dump a few folders in and see how it works. When there are errors or incomplete data, sign up for an account and fix it for everyone. It's a beautiful system, now that they finally have a decent client and a better fingerprinting algorithm.

And as an example here's a vinyl that actually contains digital content (a C=64 program):

http://musicbrainz.org/release/be404755-21ee-46aa- a945-49cc23662b53.html

It sounds as though what you really want is MusicBrainz, which creates acoustic fingerprints, rahter than relying on the filename. It worked for 95% of my music, and for the rest, I had no trouble looking up the cd details and adding it myself.

Tagging MP3s: Musicbrainz has projects to automatically tag MP3s with metadata on track name, album, etc. As for tagging them with mood: good luck; however All Music Guide has been working on this sort of thing for years; see also Last.fm. Integrating these into a desktop would be nice, though your comparison to "Microsoft tunnel vision" is quite harsh seeing as open source desktops have long had features that Windows sorely lacks, such as transparent SSH file transfers, thumbnailing of PDFs and other non-photo documents, and viewers for multiple file types, embedded right into the file manager.

MusicAnalysis uses up to 10 minutes of a track and examines all sorts of things. This is the secret sauce that makes MusicIP tick, and that allows the MusicIP mixer (aka MusicMagicMixer or MMM) to generate playlists of similar music. This is never going to be open sourced. Music analysis takes a while (about 80% of the file's playing time).

In order to generate a new PUID, you must analyze a track fully. Currently you have to use the MusicIP mixer in order to do this. The result of this analysis is submitted to the MusicDNS service and is used by the MusicDNS server to do fuzzy matching. This data is closed source, patented, and even secret (the closed source app MusicIP mixer sends the data to a closed source server, and it never sees the light of the public). The only thing that gets public is the Portable Unique IDentifier (PUID), which is a 128-bit ID of the respective analysis data on the MusicDNS server.

Which is a great reason to look at MusicBrainz.

musicbrainz automatically checks a given chunk of audio against a database of known songs -- it's designed for automated MP3 tagging, but similar tech could be used by youtube

It's not line its possible to do acoustic fingerprinting...

I looked up one of my favorite CD's -- the OST to Supergirl (1984)

Here's the proof:

MusicBrainz.org link to the 23-track SILVA AMERICA release.

'Decoded' affiliate link at Amazon.com.

Non-profit or not, musicbrainz.org 'get's paid' if you buy CDs via the amazon.com affiliate links imbedded in their site.

Slashdot CAPTCHA: squash

Ironic that a .org like slashdot has to show ads to pay the bills but that is pretty much the only way they can stay on the net. I doubt their 'subscription model' makes any sort of dent in their bandwidth bills.

I looked up one of my favorite CD's -- the OST to Supergirl (1984)

Here's the proof:

MusicBrainz.org link to the 23-track SILVA AMERICA release.

'Decoded' affiliate link at Amazon.com.

Non-profit or not, musicbrainz.org 'get's paid' if you buy CDs via the amazon.com affiliate links imbedded in their site.

Slashdot CAPTCHA: squash

Ironic that a .org like slashdot has to show ads to pay the bills but that is pretty much the only way they can stay on the net. I doubt their 'subscription model' makes any sort of dent in their bandwidth bills.

MusicBrainz is the future.

I really don't think the metadata is the issue. MusicBrainz actually supports multiple artists and many other features to enable deep metadata about music. We also have standards that support multiple pieces of identical metadata - FLAC and OGG explicitly define it, as do ID3 2.4 and, to a lesser extent, 2.3.

The problem in my view is the tag libraries and the players. UI for handling multiple artists or other information is usually abysmally poor. Winamp kind of supports it - you can search for a secondary artist, but that artist doesn't show up in the interface. I should check iTunes support more thoroughly, but I'm not expecting much better.

Some libraries for reading tags understand multiple values, but others don't. The ones that do invariably get used by lazy programmers who do things like $track->artist[0] instead of a proper for loop.

I see a python wrapper around a C++ lib and I'm not sure what the analyser is doing. C++ really isn't my thing, what I can tell you is that there's nothing clever that couldn't be easily defeated if this is the fingerprinting code.

is http://musicbrainz.org/. It's an open source music fingerprinting project that can, for example, take a hard disk full of untagged MP3s, and tag them all up fairly accurately. It's free, and it works. I *think* it uses the same engine as the UK Shazam! mobile phone service where you could ring a premium rate phone number in a club, hold your mobile up near the speaker for 15 seconds, and it'd text you back the track/artist details seconds later.
Presumably it'd be trivial for Myspace to run this in the background on the boxes where they keep user audio content...

(And no, this doesn't count. And I know CDDB doesn't have that info either.)

CDDB and FreeDB are old news. MusicBrainz is by far superior. It accounts for different release years, different formats, multiple artists, compilation albums, etc. "Why would I need to use your site? What's wrong with FreeDB?".

I'm not affiliated, just another happy user.

CDDB and FreeDB are old news. MusicBrainz is by far superior. It accounts for different release years, different formats, multiple artists, compilation albums, etc. "Why would I need to use your site? What's wrong with FreeDB?".

I'm not affiliated, just another happy user.

Most FreeDB/CDDB clients can access MusicBrainz through a CDDB gateway: http://musicbrainz.org/doc/CddbGateway

The "hashing" algorithm you are refering to is called MusicBrainz. I just started using amarok recently, and it works pretty well, although it usually gives a few unrelated choices.

Considering the poor state of many of the Freedb entries, is that data really useful? I've been volunteering with the MusicBrainz project since October and I've found the data at Freedb to be a complete mess. MusicBrainz users can use Freedb to import albums so that we don't have to re-enter things into MusicBrainz by hand, but with so many duplicate and poorly edited entries (typos, etc) I'm wondering if it's worth it to even keep the data.

MusicBrainz is a better designed system. It's not limited to the archaic interface and design of the old CDDB system. It has interfaces that programmers can use to retrieve the same kind of data that they get from Freedb. The site also has a system in place for editing of entries and peer review of changes. I think it's a better solution, although I'm biased because of my involvement and interest with the project.