Installation¶
Requirements¶
- Python 3.10 or newer
- FFmpeg and FFprobe available on
PATH libsndfile, usually installed automatically withsoundfilewheels on common platforms
Install FFmpeg¶
macOS:
brew install ffmpeg
Ubuntu or Debian:
sudo apt-get update
sudo apt-get install ffmpeg
Check that both binaries are available:
ffmpeg -version
ffprobe -version
Install The Package¶
For local development:
make install
This runs:
pip install -e ".[dev]"
pre-commit install
For installation from PyPI:
pip install audio-prep-pipeline
Install directly from GitHub:
pip install "audio-prep-pipeline @ git+https://github.com/nattkorat/audio-prep-pipeline.git"
For editable local usage without developer tools:
pip install -e .
For documentation work:
pip install -e ".[docs]"
Chunking Dependencies¶
The standard package install includes torch, silero-vad, and tqdm, so
both audio-prep convert and audio-prep chunk are ready after:
pip install audio-prep-pipeline
The first chunking run loads Silero from the installed silero-vad package.
If Silero cannot load in an offline environment, pass --allow-energy-fallback
to use the lower-quality offline energy detector.