Fix Python Librosa NoBackendError When Reading an Audio – Librosa Tutorial

When we are using python librosa to read an audio, we may get this error: NoBackendError. In this tutorial, we will introduce you how to fix it.

This error looks like:

How to fix this error?

We can read this audio using soundfile or scipy.io.wavfile.read().

Here are two tutorials:

The Difference Between scipy.io.wavfile.read() and librosa.load() in Python – Python Tutorial

Understand librosa.load() is Between -1.0 and 1.0 – Librosa Tutorial

Then, we can save audio data to pcm and read it by librosa again.

Here is an solution.

import librosa
import soundfile as sf

def save_file(audio_data, wave_file, sr = 8000):
    sf.write(wave_file, audio_data, sr, 'PCM_16')

file = "test.mp3"

try:
    data, sr = librosa.load(file, sr = 8000, mono=True)
except Exception as e:
    audio_data, samplerate = sf.read(file, dtype="float32")
    save_file(audio_data, file, sr = samplerate )

import librosa
import soundfile as sf

def save_file(audio_data, wave_file, sr = 8000):
    sf.write(wave_file, audio_data, sr, 'PCM_16')

file = "test.mp3"

try:
    data, sr = librosa.load(file, sr = 8000, mono=True)
except Exception as e:
    audio_data, samplerate = sf.read(file, dtype="float32")
    save_file(audio_data, file, sr = samplerate )

In this solution, we read test.mp3 file using soundfile, then save it to new format.

Finally, we can use librosa to read test.mp3 successfully.