Skip to content

Python bindings for Chromaprint acoustic fingerprinting and the Acoustid Web service

License

Notifications You must be signed in to change notification settings

beetbox/pyacoustid

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Chromaprint and Acoustid for Python

Chromaprint and its associated Acoustid Web service make up a high-quality, open-source acoustic fingerprinting system. This package provides Python bindings for both the fingerprinting algorithm library, which is written in C but portable, and the Web service, which provides fingerprint lookups.

Installation

This library works with Python 2 (2.7+, possibly also 2.6) and Python 3 (3.3+).

First, install the Chromaprint fingerprinting library by Lukáš Lalinský. (The library itself depends on an FFT library, but it's smart enough to use an algorithm from software you probably already have installed; see the Chromaprint page for details.) This module can use either the Chromaprint dynamic library or the fpcalc command-line tool, which itself depends on libavcodec. If you use fpcalc, either ensure that it is on your $PATH or set the FPCALC environment variable to its location.

Then you can install this library from PyPI using pip:

$ pip install pyacoustid

This library uses audioread to do audio decoding when not using fpcalc and requests to talk to the HTTP API (pip should automatically install these dependencies).

Running

You can run the included demonstration script, aidmatch.py, to test your installation:

$ python aidmatch.py mysterious_music.mp3

This will show the top metadata match from Acoustid's database. The script uses audioread to decode music, so it should transparently use a media library available on your system (GStreamer, FFmpeg, MAD, or Core Audio).

Using in Your Code

The simplest way to use pyacoustid to identify audio files is to call the match function:

>>> import acoustid
>>> for score, recording_id, title, artist in acoustid.match(apikey, path):
>>>     ...

This convenience function uses audioread to decode audio and parses the response for you, pulling out the most important track metadata. It returns in iterable over tuples of relevant information. Everything happens in one fell swoop. There are also a number of "smaller" functions you can use to perform parts of the process:

  • fingerprint(samplerate, channels, pcmiter): Generate a fingerprint for raw audio data. Specify the audio parameters and give an iterable containing blocks of PCM data.
  • fingerprint_file(path): Using either the Chromaprint dynamic library or the fpcalc command-line tool, fingerprint an audio file. (You can use force_fpcalc to use only the latter.) Returns a pair consisting of the file's duration and its fingerprint.
  • lookup(apikey, fingerprint, duration): Make a request to the Acoustid API to look up the fingerprint returned by the previous function. An API key is required, as is the length, in seconds, of the source audio. Returns a parsed JSON response.
  • parse_lookup_result(data): Given a parsed JSON response, return an iterator over tuples containing the match score (a float between 0 and 1), the MusicBrainz recording ID, title, and artist name for each match.
  • compare_fingerprints(a, b): Compare two fingerprints produced by fingerprint_file, returning a similarity score.

The module internally performs thread-safe API rate limiting to 3 queries per second whenever the Web API is called, in accordance with the Web service documentation.

If you're running your own Acoustid database server, you can set the base URL for all API calls with the set_base_url function.

Calls to the library can raise AcoustidError exceptions of two subtypes: FingerprintGenerationError and WebServiceError. Catch these exceptions if you want to proceed when audio can't be decoded or no match is found on the server. NoBackendError, a subclass of FingerprintGenerationError, is used when the Chromaprint library or fpcalc command-line tool cannot be found.

Version History

1.3.0
Multiple artists are joined by join phrases, as displayed on the web page. A new (pure-Python) function compares two Chromaprint fingerprints for similarity. Search more shared-library locations for the Chromaprint library on Windows.
1.2.3
Fix the signedness of numbers returned from chromaprint_decode_fingerprint.
1.2.2
Fix a regression in the previous version that caused a KeyError crash when calling submit.
1.2.1
The meta parameter to some API functions can now be a list (instead of just a single string).
1.2.0
Add a force_fpcalc option to fingerprint_file and match. Avoid leaving a dangling socket after communicating with the server. Fix a crash when passing a memoryview object to the fingerprinter. API requests can now optionally time out. More reliably find the library on Windows on Python 3.8. Add a hash_fingerprint function to the low-level library.
1.1.7
Include a LICENSE file.
1.1.6
In submission, avoid an error on non-integer durations. A new function, get_submission_status, abstracts the API endpoint for monitoring submissions using the (new) result from the submit function.
1.1.5
Fix compatibility with Python 3 in the submit function. Errors in submit are now also handled correctly (i.e., they raise an informative WebServiceError instead of a TypeError).
1.1.4
Fix an error on versions of the fpcalc tool that report the duration as a fractional number.
1.1.3
Accept bytearray objects in addition to other bytes-like types.
1.1.2
Fix a possible crash on Unicode text in Python 2 in a non-Unicode locale. Look for version "1" of the Chromaprint shared library file.
1.1.1
Fix a possible setup error on Python 3 (thanks to Simon Chopin).
1.1.0
Include fpcalc.py script in source distributions. Add Python 3 support (thanks to Igor Tsarev).
1.0.0
Include fpcalc.py, a script mimicking the fpcalc program from the Chromaprint package. Handle a UnicodeDecodeError raised when using the fpcalc backend on Windows with Unicode filenames. Standard error output from fpcalc is suppressed.
0.7
Properly encode Unicode parameters (resolves a UnicodeEncodeError in fingerprint submission). Parse all recordings for each Acoustid lookup result.
0.6
Add a new function, fingerprint_file, that automatically selects a backend for fingerprinting a single file.
0.5
Fix response parsing when recording has no artists or title. Fix compatibility with Python < 2.7. Add specific NoBackendError exception.
0.4
Fingerprinting can now fall back to using the fpcalc command-line tool instead of the Chromaprint dynamic library so the library can be used with the binary distributions (thanks to Lukáš Lalinský). Fingerprint submission (thanks to Alastair Porter). Data chunks can now be buffers as well as bytestrings (fixes compatibility with pymad).
0.3
Configurable API base URL. Result parser now generates all results instead of returning just one. Find the chromaprint library on Cygwin. New module names: chromaprint and acoustid (no package).
0.2
Compress HTTP requests and responses. Limit audio decoding to 120 seconds. Return score from convenience function.
0.1
Initial release.

Credits

This library is by Adrian Sampson. Chromaprint and Acoustid are by Lukáš Lalinský. This package includes the original ctypes-based bindings written by Lukáš. The entire library is made available under the MIT license. pyacoustid was written to be used with beets, which you should probably check out.