NAME
Audio::Ofa::Util - Retrieve audio fingerprints and metadata for unknown
audio files
SYNOPSIS
This module tries to make retrieving audio fingerprints and metadata for
unknown audio files as easy as possible. It interfaces with the modules
Audio::Ofa and WebService::MusicBrainz, provides a simple LWP based
interface to the MusicDNS library, and can make use of
Audio::Extract::PCM to read some popular music formats.
The most comprehensive way to use this is to start with a (possibly
untagged) file name and get full metadata:
my $util = Audio::Ofa::Util->new(filename => 'song.ogg');
my @tracks = $util->musicbrainz_lookup or die $util->error;
for (@tracks) {
print 'Artist: ', $_->artist, "\n";
print 'Title: ', $_->title, "\n";
print 'Track: ', $_->track, "\n";
print 'Album: ', $_->album, "\n\n";
}
To create an audio fingerprint:
my $util = Audio::Ofa::Util->new(filename => 'song.ogg');
$util->analyze_file or die $util->error;
print $util->fingerprint, "\n";
To create a fingerprint and look it up at MusicDNS:
my $util = Audio::Ofa::Util->new(filename => 'song.ogg');
$util->musicdns_lookup or die $util->error; # calls analyze_file implicitly
print $util->artist, ' - ', $util->title, "\n";
To look up a known fingerprint at MusicDNS (you need the length of the
song, too):
my $util = Audio::Ofa::Util->new(fingerprint => $fp, duration => $millisecs);
The overall process goes like this:
* We create an audio fingerprint, which stores some
characteristics of a recording in a rather small amount of data.
This is what libofa (and the Perl binding in Audio::Ofa) does.
This module (Audio::Ofa::Util) faciliates this with
"analyze_file" by allowing to fingerprint some widely used music
formats and storing the results so they can be used for the next
steps:
* The audio fingerprint is submitted to the MusicDNS web service.
Using a proprietary fuzzy algorithm and their database, they
determine which song we have at hand. MusicDNS returns some
metadeta: The artist, the song title, and a PUID. This "portable
unique identifier" is an arbitrary index into their database and
is unique for every recording of a given song.
Note that while libofa's audio fingerprints may change after
transformations of a recording (such as lossy audio compression
or radio transmission), the fuzzy algorithm will (ideally) still
find the same PUID.
* Because we usually want to know more than the artist and title,
we look up the PUID in a second Web Database called MusicBrainz.
It provides us with all desired metadata such as all the albums
the song has appeared on in this particular version, and the
respective track numbers.
This module provides a basic MusicBrainz PUID lookup through
"musicbrainz_lookup". If you want to know even more (such as
members of the band and the previous bands of those members),
you can use WebService::MusicBrainz, to which this module
provides an easy frontend.
ACCESSORS
filename
See "analyze_file".
fingerprint, duration
See "analyze_file" and "musicdns_lookup".
client_id, client_version, metadata, bitrate, extension, artist, title,
album, track, genre, year, puids
See "musicdns_lookup".
Note that puids accesses an array reference. If it is not defined or not
set, it means that no PUID has been looked up yet. If it is an empty
array, it means that no PUIDs were found.
error
Description of the last error that happened.
METHODS
new
Constructor. Accepts key-value pairs as initializers for all of the
fields, c.f. "ACCESSORS", but currently only the following calls make
sense:
Audio::Ofa::Util->new(filename => $filename);
Audio::Ofa::Util->new(fingerprint => $fp, duration => $dur);
Audio::Ofa::Util->new(puid => $puid);
analyze_file
This creates an Audio Fingerprint of a sound file. The audio file is
read using Audio::Extract::PCM, which currently uses the extarnal "sox"
program and supports encodings such as MP3, Ogg/Vorbis and many others.
You must set "filename" before calling this method.
The fingerprint is calculated by "ofa_create_print" in Audio::Ofa, and
the "fingerprint" field of the object will be set. Additionally, the
"duration" (in milliseconds) and the "extension" will be set to the
values provided by the file name.
In case of an error, an empty list is returned and the error message can
be retrieved via "error". Otherwise, a true value will be returned.
musicdns_lookup
This looks up a track at the MusicDNS web service.
To do a fingerprint lookup, the keys "fingerprint" and "duration" must
be present, where duration is the length of the song in milli seconds.
Additionally, the following fields (defaults in parentheses) will be
sent to the MusicDNS service:
client_id (hardcoded client id), client_version (module name and
version), fingerprint, metadata (1), bitrate (0), extension ("unknown"),
duration, artist ("unknown"), title ("unknown"), album ("unknown"),
track (0), genre ("unknown"), year (0).
To do a fingerprint lookup, "fingerprint" and "duration" must have been
set (can be given to "new"), where "duration" is the song length in
milli seconds.
If "fingerprint" hasn't been set, "analyze_file" is called implicitly.
client_id defaults to a hard-coded Client ID. You can get your own from
http://www.musicip.com.
You should set as much of the above-mentioned metadata (like artist,
etc.) as you have available, because the MusicDNS terms of service
require this in order to help clean errors in their database.
In the case of an error, "musicdns_lookup" returns an empty list and the
error message can be retrieved with the "error" method.
In the case of success, "musicdns_lookup" sets the fields "puids" to the
found PUIDs, and sets the fields "artist" and "title" to the first of
the found values, and returns a true value. In list context, it returns
a list of objects which have "artist", "title" and "puid" methods.
musicbrainz_lookup
This looks up a PUID at MusicBrainz. The PUID can come from a call to
"musicdns_lookup". In fact this is implicitly done if there is no PUID
stored in the object (cf. "SYNOPSIS").
This returns a list of WebService::MusicBrainz::Response::Track objects
on success, or the first of them in scalar context. Otherwise it returns
an empty list and the error message can be retrieved via the "error"
method.
This method returns a list of tracks or the first track in scalar
context. The tracks are represented as objects that are guaranteed to
have the methods "artist", "title", "album", "track" and "wsres", where
the latter is an WebService::MusicBrainz::Response::Track object, and
the four former return values that have been retrieved from that object
for your convenience.
In the case of an error, an empty list is returned and the error can be
returned via the "error" method.
SEE ALSO
* MusicBrainz::Client - A client for the old MusicBrainz web
service
* MusicBrainz::TRM - Obsolete TRM-based audio fingerprinting
library
* tunepimp - C library which does pretty much everything that this
module does. It even includes Perl bindings, but as of this
writing, they don't compile in the current tunepimp version and
only support the old TRM fingerprints.
* <http://www.musicdns.org> - Web site of the MusicDNS web service
as provided by MusicIP
* <http://www.musicbrainz.org> - Web site of MusicBrainz
* <http://wiki.musicbrainz.org/HowPUIDsWork> - How PUIDs work
LICENSE
This module is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License (GPL) as published by
the Free Software Foundation (http://www.fsf.org/); either version 2 of
the License, or (at your option) any later version.
The GPL, which is quite restrictive (when compared to LGPL or Artistic),
seems to be necessary because of libofa's licenses, but IANAL and if you
need a license change please contact me.
Please note that in addition to the license which allows you to use this
software, the MusicDNS web service has its own terms of service. The
most important fact is that you can use it for free unless you use it
commercially. See <http://www.musicdns.org> for more information. You
are encouraged to register your own client id (for free) if you build a
client on top of this module.
AUTHOR
Christoph Bussenius (pepe at cpan.org)
Please mention the module's name in the subject of your mails so that
they will not be lost in the spam.
If you find this module useful I'll be glad if you drop me a note.