Switch speed perturbation to use SOX

ryanleary / patter

speech-to-text in pytorch

MIT License

80 stars 17 forks source link

Open ryanleary opened 6 years ago

ryanleary commented 6 years ago

Librosa's implementation seems to cause issues despite being faster.

ryanleary commented 6 years ago

Potentially look into something like audiotsm (https://audiotsm.readthedocs.io/en/latest/) and benchmark.