Skip to content

Latest commit

 

History

History
23 lines (18 loc) · 943 Bytes

README.md

File metadata and controls

23 lines (18 loc) · 943 Bytes

wsj0-convert

A Python script to convert the WSJ0 speech corpus to more friendly file formats.

Requirements

  • sph2pipe in PATH. Get it from here.
  • ffmpeg in PATH (not required if using the --no-flac option)

Usage

Simply run:

python wsj0_convert.py <path-to-WSJ0> <output-dir>

This will create an audio directory inside <output-dir> containing all the audio files in .flac format, or in .wav format if using the --no-flac option. The audio files are organized in sub-directories by speakers.

Notes

  • Only .wv1 files are converted. .wv2 files are skipped.
  • The original folder structure is not respected. Output files are organized in sub-directories by speaker.
  • Total output size if 3.9 GB when using .flac format.
  • For extra speaker information (e.g. gender), see here.