Acoustic fingerprinting is a technique for identifying songs from the way they “sound” rather from their existing metadata. That means that beets’ autotagger can theoretically use fingerprinting to tag files that don’t have any ID3 information at all (or have completely incorrect data). This plugin uses an open-source fingerprinting technology called Chromaprint and its associated Web service, called Acoustid.
Turning on fingerprinting can increase the accuracy of the autotagger—especially on files with very poor metadata—but it comes at a cost. First, it can be trickier to set up than beets itself (you need to set up the native fingerprinting library, whereas all of the beets core is written in pure Python). Also, fingerprinting takes significantly more CPU and memory than ordinary tagging—which means that imports will go substantially slower.
If you’re willing to pay the performance cost for fingerprinting, read on!
First, install pyacoustid itself. You can do this using pip, like so:
$ pip install pyacoustid
Then, you will need to install Chromaprint, either as a dynamic library or in the form of a command-line tool (fpcalc).
Installing the Binary Command-Line Tool¶
The simplest way to get up and running, especially on Windows, is to download the appropriate Chromaprint binary package and place the fpcalc (or fpcalc.exe) on your shell search path. On Windows, this means something like C:\\Program Files. On OS X or Linux, put the executable somewhere like /usr/local/bin.
Installing the Library¶
On OS X and Linux, you can also use a library installed by your package manager, which has some advantages (automatic upgrades, etc.). The Chromaprint site has links to packages for major Linux distributions. If you use Homebrew on Mac OS X, you can install the library with brew install chromaprint.
You will also need a mechanism for decoding audio files supported by the audioread library:
- OS X has a number of decoders already built into Core Audio, so there’s no need to install anything.
- On Linux, you can install GStreamer for Python, FFmpeg, or MAD and pymad. How you install these will depend on your distribution. For example, on Ubuntu, run apt-get install python-gst0.10-dev. On Arch Linux, you want pacman -S gstreamer0.10-python. If you use GStreamer, be sure to install its codec plugins also.
- On Windows, try the Gstreamer “WinBuilds” from the OSSBuild project.
To decode audio formats (MP3, FLAC, etc.) with GStreamer, you’ll need the standard set of Gstreamer plugins. For example, on Ubuntu, install the packages gstreamer0.10-plugins-good, gstreamer0.10-plugins-bad, and gstreamer0.10-plugins-ugly.
Once you have all the dependencies sorted out, you can enable fingerprinting by editing your configuration file. Put chroma on your plugins: line. With that, beets will use fingerprinting the next time you run beet import.
You can also use the beet fingerprint command to generate fingerprints for items already in your library. (Provide a query to fingerprint a subset of your library.) The generated fingerprints will be stored in the library database. If you have the import.write config option enabled, they will also be written to files’ metadata.
You can help expand the Acoustid database by submitting fingerprints for the music in your collection. To do this, first get an API key from the Acoustid service. Just use an OpenID or MusicBrainz account to log in and you’ll get a short token string. Then, add the key to your config.yaml as the value apikey in a section called acoustid like so:
acoustid: apikey: AbCd1234
Then, run beet submit. (You can also provide a query to submit a subset of your library.) The command will use stored fingerprints if they’re available; otherwise it will fingerprint each file before submitting it.