Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


Bard Music Manager - A database to manage your music, find duplicates and fix tags


Right now there are only prebuilt packages for openSUSE Leap and Tumbleweed. In those cases, you can add my build repository:

sudo zypper ar obs://home:alarrosa:bard bard-repository

Once the repository is added, you can install bard with:

sudo zypper in bard

Build from sources

Install build dependencies

First, some dependencies have to be installed in order to build bard:

sudo zypper in libboost_python3-devel "pkgconfig(libavcodec)" "pkgconfig(libavformat)" "pkgconfig(libswresample)" "pkgconfig(libavutil)" python3-pyacoustid python3-mutagen python3-Pillow python3-numpy python3-dbus-python python3-SQLAlchemy python3-pydub python3-SQLAlchemy-Utils python3-alembic python3-paramiko
sudo apt-get install build-essential python3-dev libavcodec-dev libavformat-dev libswresample-dev libavutil-dev python3-acoustid python3-mutagen python3-numpy python3-sqlalchemy python3-setuptools libboost-python3-dev

If you want to use the web interface (which is still in early stages of development, and thus not ready for real usage):

sudo zypper in python3-Werkzeug python3-Flask python3-Flask-Cors python3-Flask-Login python3-Jinja2 python3-bcrypt

If you want bard to do high level analysis of songs (extracting tonal, rhythm, genre and other features from an audio analysis) you need to install the latest essentia version and its highlevel models. I've prepared openSUSE packages in the bard repository. They can be installed with:

sudo zypper in essentia-python essentia-models

If you intend to build the internal bard tests (only recommended if you plan to contribute to bard), then you should also install:

sudo zypper in libboost_program_options-devel
sudo apt-get install libboost-program-options-dev

Building bard

And now we're ready to build and install bard:

python3 build
python3 install


Before using bard, you need to configure it. Configuring it is very easy and mostly implies telling it a list of directories where you have your music files, so bard can scan them and extract information. Note that bard will never modify or write over your music files or directories. All it does is read them, extract tags, perform an audio analysis and write everything into its database.

You can run bard init to create a basic example config file with a basic structure that you can edit to configure bard instead of having to do it from scratch.

The example file will be written to the default config file location at ~/.config/bard. Edit it and set at least the correct values for the music_paths config option. It should be set to one or more directories where you have music collection. Bard will search for music recursively on directories under the ones you specify. Note that Bard tries to help you to organize your music collection and keep it clean, so it will complain if it finds files it cannot recognize. In those cases you might want to move the files out of your music collection directory or in some cases, even remove the files. If you want to keep them, you can add glob patterns to the ignore_files config option. Files matching those globs will be ignored by Bard.

Another important aspect you might want to decide is what database Bard should use. By default a sqlite database will be used and you don't have to worry about configuring anything, but if you have the possibility of using a postgresql database, that's preferred. Refer to the configuring postgresql documentation for information on how to do that.

Note that there's currently no migration process from one database to another that means you might want to choose the database in an early stage so you don't have to process all your collection again if you change to use another database.


bard update

will update the internal database with new and modified files from the directories into the configuration file. The music files will be read and information like format, embeeded tags, audio quality, file hash, decoded audio hash, audio fingerprint, etc. will be extracted and stored in the database.

Bard will assign each song found an ID that identifies the song in the database.

bard ls Dire%Straits

will list all songs having Dire.*Straits into their path (note that SQL's LIKE wildcards are used here).

bard play Glass

will play (using mpv) all songs containing "Glass" into their path

Note that for most bard commands, you can use "search expressions", full paths to a file, or the ID given to a song by bard.

bard process-songs

While bard update searches for new/removed songs and extracts basic metadata about them, the process-songs command is a lenghtier process that performs music analysis. It's composed on two kinds of analysis that are performed in two steps and can be run independently with the find-audio-duplicates and analyze-songs commands.

The first step, equivalent to find-audio-duplicates, compares the audio fingerprint of a song to all other existing fingerprints in the database thus finding similar songs even if they have different format, audio quality or (in some cases) even if they are different versions of the same song. The similarity between songs is stored in the database and can be seen with the info command when showing information about a song (which shows which songs are similar) and is also used by the compare-dirs and compare-songs commands (which compares directories or individual songs).

The second step, equivalent to analyze-songs, uses the essentia library to analyze the songs and extract audio features such as tonality or rhythm information and also uses highlevel classifiers to extract a probability of a song to be instrumental or having voices (in that case, also, if there's a male or female voice), the probability of a song to be acoustic, electronic, bright, dark, aggresive, relaxed and even several automatic genre classifiers.

Note that this second step requires a working installation of the essentia library, python bindings and compatible classifier models. If you installed Bard using the openSUSE packages, this should have been installed automatically.

As a shortcut, you can also use bard update --process which runs the update command and after it finishes, it runs the process-songs command.

bard info Maximizing%the%audience

Will show all information about any song containing "Maximizing.*the.*audience" into its path.

bard find-audio-duplicates

Will take a while and analyze all songs, comparing the audio signatures and storing which songs are similar into the database. To see if one song has any duplicate, just use the info command. bard find-audio-duplicates should be run every time new songs are added to a collection. Note that using bard process-songs is recommended instead as it does all audio processing at once.

bard compare-dirs -s directory1 directory2

Will compare two directories and report if the songs in the first directory are a subset of the songs in the second directory. Being a subset means that files are similar and having equal or less quality (for example if directory1 and directory2 contains the same songs, but the ones in directory1 are encoded in mp3 128kbps while the songs in directory2 are encoded in mp3 320 kbps).

Note that bard will never remove files. If you find duplicated files, you can remove them and then run bard update to let bard know that some files no longer exist.

bard compare-files file1 file2

This command compares two songs, extracting the audio fingerprint and comparing them. Since the command extracts the audio fingerprint instead of using what's stored in the database, it works with files that haven't been added to the database yet, even for files that are outside the music_paths directories. Alternatively, there's a compare-songs command that uses the fingerprint stored in the database and thus it's much faster but can only be used for songs already added to the database.

Recommended workflow to organize your music collection

The recommended workflow to organize your music is this:

  1. Put all your unorganized music in one or more directories and add those directories to the music_paths config option in Bard.
  2. Choose one empty directory where you'll have "correctly organized music", add that directory to the music_paths config option in Bard and also put it in the musicbrainz_tagged_music_paths config option. Finally open Picard and in the Options dialog, in the File Naming section, set the Destination directory to that directory too.
  3. You can now start using Bard adding songs to its collection, analyzing and playing them and comparing different directories and song files.
  4. Install the Bard plugin and file naming script in Picard.
  5. Go over the albums in the unorganized music directories adding MusicBrainz tags with Picard and saving them. Picard will move them to the organized folder. I cannot stress enough the importance of correct songs having correct MusicBrainz tags. Note that this for example disambiguates Artists who are called the same and even different versions of songs, not only live or studio versions but also really different versions of songs that some artists sometime record.
  6. After organizing files, you have to run bard update so bard notices the files were moved. Note that Bard will usually detect that files were moved and just update their location keeping their ID and extracted analysis information. Bard will detect moved files even if their tags or cover art are modified since it checks the audio hash for this.
  7. The Bard plugin for Picard will write a hidden .artist_mbid file to artist directories. On those directories, you can optionally place a file named "artist.jpg" with the corresponding artist image. This will be processed by the update-musicbrainz-artists command so the web interface will use that image for that artist. It's obviously recommended to use those image files.

You can use bard stats to get statistics on your collection, including the percentage of songs correctly organized (having MusicBrainz tags)

MusicBrainz integration

Bard can import information from the MusicBrainz database into your local database. To do this there are two command that have to be run periodically.

bard mb-update

This will download the latest MusicBrainz database dumps from the MusicBrainz servers and extract them. Note that the MusicBrainz database is huge so at the time of writing this, the mb-update command requires ~17GB in your home directory (~3.7GB to download and ~13.3GB after uncompression). You can run this command as many times as you want (it'll just update the downloaded and extracted files) but note that the database dumps are only updated in MusicBrainz servers every 3 or 4 days, so you don't need to execute it every day.

bard mb-import

Since the database is so big, we don't want to import all of it as Bard only will use information related to the music you have in the collection. Using the mb-import command, Bard will first gather which information is useful to it and then import only that. Note that only the information about songs with correct MusicBrainz tags will be imported (which makes it important to use Picard and organize your music correctly). As an example, note that Bard will not only import information about the main artist a song is credited to, but also information about other artists related to the song (guest performers, for example) and information about the release/release group (such as release country, date, etc.).

Note that the web interface currently only shows songs from your collection having MusicBrainz imported data. Songs in the "unorganized music directories" can be seen from the command line, but the web interface is currently restricted to organized music.

Doing Backups of your music collection

Currently backups are only done to a different computer which can be accessed using ssh. Check the backup documentation for more information.

The web interface

First, remember to periodically run bard update-musicbrainz-artists to update the artist images located next in artist folders as explained in the recommended workflow above.

Then, you want to set a password to your user. You can do this by running bard passwd. Bard will request the password and store it (hashed) in the database. When logging into the web interface, use your username and password to authenticate. If you want to use a different username from your unix username, set the username config option to that and run bard passwd yourusername to specify the username you want to change the password to.

bard web

If you have Bard installed in your system, you can start the web interface by running bard web. Then you can point your web browser to the hostname used in the configuration and port 5000 (by default, unless changed in the configuration).

The web interface is kind-of usable. You can browse your artists, see their release groups, releases and tracks within the releases. You can also play songs, create playlists, drag and drop songs from releases to your playlists, set ratings and perform searches (on path names, which usually include the artist names the release names as well as the track names). But also there are many things you will be missing. Most noticeable missing features in the web ui are: Removing songs from playlists, reordering songs within a playlist, removing playlists, view song information, etc. I also plan to support playing music through Sonos and Chromecast devices at some point but that doesn't work yet neither.

About music sources

Please, always use music from legal sources and support the artists.


Bard is distributed under the GPL v3.0 license. The web interface uses jQuery, which is licensed under a MIT license.


Bard Music Manager - A database to manage your music, find duplicates and fix tags







No packages published