Handle librosa version issue

[KarlosMuradyan]

There is trouble when loading .mp3 files with librosa. The versions of the libraries that librosa uses is below. The problem most probably comes either from librosa or audioread, which is used in librosa.

.wav files on the other hand, load and process properly.

INSTALLED VERSIONS
------------------
python: 3.7.3 (default, Jun 24 2019, 04:54:02) 
[GCC 9.1.0]

librosa: 0.7.1

audioread: 2.1.8
numpy: 1.17.0
scipy: 1.3.1
sklearn: 0.21.3
joblib: 0.13.2
decorator: 4.4.0
six: 1.12.0
soundfile: 0.10.2
resampy: 0.2.2
numba: 0.46.0

[bezirganyan]

Librosa seems to work fine with mp3 files in version 0.7.2 . It throws a warning:

.../site-packages/librosa/core/audio.py:161: UserWarning: PySoundFile failed. Trying audioread instead.
  warnings.warn('PySoundFile failed. Trying audioread instead.')

but loads the file just fine. From an issue in librosa, it tries to use libsndfile first, and if that fails, it will fall back on the audioread package.

Can you verify whether the problem persists in the newer version?

python 3.7.5

librosa==0.7.2

audioread==2.1.8
cffi==1.13.2
decorator==4.4.1
joblib==0.14.1
llvmlite==0.31.0
numba==0.47.0
numpy==1.18.1
pkg-resources==0.0.0
pycparser==2.19
resampy==0.2.2
scikit-learn==0.22.1
scipy==1.4.1
six==1.14.0
SoundFile==0.10.3.post1

[KarlosMuradyan]

I can confirm that the problem still exists for me even after upgrading the librosa package. Now the versions look like this:

INSTALLED VERSIONS
------------------
python: 3.7.3 (default, Jun 24 2019, 04:54:02) 
[GCC 9.1.0]

librosa: 0.7.2

audioread: 2.1.8
numpy: 1.17.0
scipy: 1.3.1
sklearn: 0.21.3
joblib: 0.13.2
decorator: 4.4.0
six: 1.12.0
soundfile: 0.10.2
resampy: 0.2.2
numba: 0.46.0

I started a discussion in librosa google forum. Please look at this link for the details and the sample audio files.