Search code examples
pythonlibrosa

How to split audio (side a and side b) using librosa

I'm trying to use librosa to split a stereo audio file into separate channels. I have this code:

import librosa

audio, sr = librosa.load('trumpet.wav')

canal_esquerdo, canal_direito = librosa.effects.split_stereo(audio)

librosa.output.write_wav('canal_esquerdo.wav', canal_esquerdo, sr)
librosa.output.write_wav('canal_direito.wav', canal_direito, sr)

but I get an error:

Traceback (most recent call last):
  File "audio.py", line 7, in <module>
    canal_esquerdo, canal_direito = librosa.effects.split_stereo(audio)
AttributeError: module 'librosa.effects' has no attribute 'split_stereo'

What is wrong with the code? Is it possible to split the audio using librosa? If so, how?

Solution

  • The error message means exactly what it says: .split_stereo isn't provided here, so that can't be used to split the stereo file.

    To find out how to use a library for a given task, do not guess at names for functions - search for, and read, documentation.

    The relevant documentation explains:

    When working with multi-channel signals, we may choose to skip the default down-mixing step by specifying mono=False in the call to librosa.load, as in the following:

    import librosa

# Get the "high-quality" multi-channel version of
# an example track
filename = librosa.ex('trumpet', hq=True)

# Load as multi-channel data
y_stereo, sr = librosa.load(filename, mono=False)

    The resulting object now has two dimensions instead of one, with y_stereo.shape == (N_channels, N_samples). This way, we can access the first channel as y_stereo[0], the second channel as y_stereo[1], and so on if there are more than two channels.

    That is to say: librosa does not treat "stereo" as a special case; as long as mono=False is specified, loading just gives a 2d array representing however many channels there are.

    Adapting that to the original code:

    import librosa

audio, sr = librosa.load('trumpet.wav', mono=False)

librosa.output.write_wav('canal_esquerdo.wav', audio[0], sr)
librosa.output.write_wav('canal_direito.wav', audio[1], sr)