aubio

aubio is a collection of tools for music and audio analysis.

This package integrates the aubio library with NumPy to provide a set of efficient tools to process and analyse audio signals, including:

• read audio from any media file, including videos and remote streams
• high quality phase vocoder, spectral filterbanks, and linear filters
• Mel-Frequency Cepstrum Coefficients and standard spectral descriptors
• detection of note attacks (onset)
• pitch tracking (fundamental frequency estimation)
• beat detection and tempo tracking

aubio works with both Python 2 and Python 3.

Links

• module documentation
• installation instructions
• aubio manual
• aubio homepage
• issue tracker

Demos

Some examples are available in the python/demos folder. Each script is a command line program which accepts one ore more argument.

Notes: installing additional modules is required to run some of the demos.

Analysis

• demo_source.py uses aubio to read audio samples from media files
• demo_onset_plot.py detects attacks in a sound file and plots the results using matplotlib
• demo_pitch.py looks for fundamental frequency in a sound file and plots the results using matplotlib
• demo_spectrogram.py, demo_specdesc.py, demo_mfcc.py for spectral analysis.

Real-time

• demo_pyaudio.py and demo_tapthebeat.py use pyaudio
• demo_pysoundcard_play.py, demo_pysoundcard.py use PySoundCard
• demo_alsa.py uses pyalsaaudio

Others

• demo_timestretch.py can change the duration of an input file and write the new sound to disk,
• demo_wav2midi.py detects the notes in a file and uses mido to write the results into a MIDI file

Example

Use demo_timestretch_online.py to slow down loop.wav, write the results in stretched_loop.wav:

$ python demo_timestretch_online.py loop.wav stretched_loop.wav 0.92

Built with

The core of aubio is written in C for portability and speed. In addition to NumPy, aubio can be optionally built to use one or more of the following libraries:

• media file reading:

  • ffmpeg / avcodec to decode and read audio from almost any format,
  • libsndfile to read audio from uncompressed sound files,
  • libsamplerate to re-sample audio signals,
  • CoreAudio to read all media formats supported by macOS, iOS, and tvOS.

• hardware acceleration:

  • Atlas and Blas, for accelerated vector and matrix computations,
  • fftw3, to compute fast Fourier Transforms of any size,
  • Accelerate for accelerated FFT and matrix computations (macOS/iOS),
  • Intel IPP, accelerated vector computation and FFT implementation.