tts-corpus-workbench

CLI utilities for inspecting and cleaning a TTS corpus.

Install

• Requires Python 3.9+.
• Install in editable mode (ensures entry point is available): pip install -e .
• The tools use soundfile, mutagen, and pandas. If pandas is missing, add it: pip install pandas.

Usage

• Show help: tts-corpus-workbench --help
• All commands accept comma-separated audio extensions (default: .wav,.mp3,.flac,.m4a).

Compute audio hours

• Count total duration in a folder:
  tts-corpus-workbench compute-audio-hours --folder data/audio --extensions ".wav,.flac"

Find orphan audio

• Compare metadata CSV with an audio folder:
  tts-corpus-workbench find-orphan-audio --metadata metadata.csv --audio-col file --folder data/audio
• Prints counts for matches/missing/orphans; add --delete-orphan to remove files on disk that are not in metadata.

Detect acronyms

• Export acronym frequency stats from a text column:
  tts-corpus-workbench detect-acronyms --metadata metadata.csv --text-col text --output acronym_stats.csv