Skip to content

A beets plugin for obsessive-compulsive music geeks to add low and high level musical information to their songs.


Notifications You must be signed in to change notification settings


Folders and files

Last commit message
Last commit date

Latest commit



83 Commits

Repository files navigation

Test & Release & Deploy Coverage Status PyPi PyPI pyversions MIT license

Xtractor (Beets Plugin)

The beets-xtractor plugin lets you, through the use of the Essentia extractors, to obtain low and high level musical information from your songs.

Currently, the following attributes are extracted for each library item: bpm, danceability, beats_count, average_loudness, danceable, gender, is_male, is_female, genre_rosamerica, voice_instrumental, is_voice, is_instrumental, mood_acoustic, mood_aggressive, mood_electronic, mood_happy, mood_sad, mood_party, mood_relaxed, mood_mirex, mood_mirex_cluster_1, mood_mirex_cluster_2, mood_mirex_cluster_3, mood_mirex_cluster_4, mood_mirex_cluster_5


The plugin can be installed via:

pip install beets-xtractor

and activated the usual way by adding xtractor to the list of plugins in your configuration:

  - xtractor

Install the Essentia extractors

You will also need the streaming_extractor_music binary extractor from the Essentia project. Please refer to the official installation documentation for the installation procedure for your OS. If you are stuck you can use the Issue tracker. If you are compiling Essentia from source, make sure you compile it with Gaia support (--with-gaia) otherwise will not be able to use the high level models.

Download the SVM models

For computing the high-level descriptors you will need the prebuilt trained models for prediction. You need to download these from here: SVM Models. I suggest that you download the most recent beta5 version. This means that your binaries must match this version. Put the downloaded models in any folder from which they can be accessed.


All your configuration will need to go under the xtractor key. This is what your configuration should look like:

  auto: no
  dry-run: no
  write: yes
  threads: 1
  force: no
  quiet: no
  keep_output: yes
  keep_profiles: no
  output_path: /mnt/data/xtraction_data
  essentia_extractor: /mnt/data/extractors/beta5/streaming_extractor_music
        - /mnt/data/extractors/beta5/svm_models/danceability.history
        - /mnt/data/extractors/beta5/svm_models/gender.history
        - /mnt/data/extractors/beta5/svm_models/genre_rosamerica.history
        - /mnt/data/extractors/beta5/svm_models/mood_acoustic.history
        - /mnt/data/extractors/beta5/svm_models/mood_aggressive.history
        - /mnt/data/extractors/beta5/svm_models/mood_electronic.history
        - /mnt/data/extractors/beta5/svm_models/mood_happy.history
        - /mnt/data/extractors/beta5/svm_models/mood_sad.history
        - /mnt/data/extractors/beta5/svm_models/mood_party.history
        - /mnt/data/extractors/beta5/svm_models/mood_relaxed.history
        - /mnt/data/extractors/beta5/svm_models/voice_instrumental.history
        - /mnt/data/extractors/beta5/svm_models/moods_mirex.history

First of all, you will need adjust all paths. Put the path of the extractor binary in essentia_extractor and substitute the location of the SVM models with your local path under the svm_models section. Finally, set the output_path to indicate where the extracted data files will be stored. If you do not set this, a temporary path will be used. will be used.

Note on shell tilde expansion: Please note that you cannot use shell expansion on the svm_models (i.e.: do not use ~ for your home folder). The entire section of extractor_profile is passed as-is to the essentia extractor binary and it will not do tilde expansion on your paths. The rest of the path keys such as essentia_extractor and output_path are used by the plugin itself and it will take care of expanding the tilde symbol (~) to the home directory of the user running the script. care of expanding the tilde symbol (~) to the home directory of the user running the script.

By default both keep_output and keep_profile options are set to no. This means that after extraction (and the storage of the important information) the profile files used to pass to the extractors, and the json files created by the extractors will be deleted. There are various reasons you might want to keep these files. One is for debugging purposes. Another is to see what else is in these files (there is a lot) and maybe to use them with some other projects of yours. Lastly, you might want to keep these because the plugin only extracts data if these files are not present. If you store them, on a successive extraction, the plugin will skip the extraction and use these files (they are named by mb_trackid) - speeding up the process a lot.

The force option instructs the plugin to execute on items which already have the required properties.

The threads option sets the number of concurrent executions. By default this is set to 1. If you remove this option or if you set it to 0 the number of CPU cores present on your machine will be used. The extraction is quite a CPU intensive process so there might be cases when you want to limit it to just 1.

The write option instructs the plugin to write the extracted attributes to the media file right away. Note that only bpm is actually written to the media file, all the other attributes are flex attributes and are only stored in the database.

The dry-run option shows what would be done without actually doing it.

NOTE: Please note that the auto option is not yet implemented. For now you will have to call the xtractor plugin manually.


Invoke the plugin as:

$ beet xtractor [options] [QUERY...]

For a more verbose reporting use the -v flag on beet:

$ beet -v xtractor [options] [QUERY...]

The plugin has also got a shorthand xt so you can also invoke it like this:

$ beet xt [options] [QUERY...]

The following command line options are available:

--dry-run [-d]: Only show what would be done - displays the extracted values but does not store them in the library.

--write [-w]: Write the values (bpm only) to the media files.

--threads=THREADS [-t THREADS]: The number of concurrently running executions.

--force [-f]: Force the analysis of all items (skip attribute checks).

--count-only [-c]: Show the number of items to be processed and exit. Extraction will not be executed.

--quiet [-q]: Run without any output.

--version [-v]: Display the version number of the plugin. Useful when you need to report some issue and you have to state the version of the plugin you are using.

These command line options will override those specified in the configuration file.


  • If something is not working as expected please use the Issue tracker.
  • If the documentation is not clear please use the Issue tracker.
  • If you have a feature request please use the Issue tracker.
  • In any other situation please use the Issue tracker.

Other plugins by the same author


Essentia is an open-source C++ library with Python bindings for audio analysis and audio-based music information retrieval. It is released under the Affero GPLv3 license and is also available under proprietary license upon request. This plugin is just a mere wrapper around this library. Learn more about the Essentia project


Final Remarks



A beets plugin for obsessive-compulsive music geeks to add low and high level musical information to their songs.