extract_loudest_section_python

This is a Python port of the C++ version of Pete Warden's extract_loudest_section tool. Distrubuted under the same license.

The extract_loudest_section_python script processes WAV files to extract the loudest segment of a specified length and saves it to a new file. It uses a sliding window approach to efficiently compute the sum of squares of the samples, which is used as a measure of volume.

The main reason for a Python port is that the original repository is an OSX only build (see Pete's blog for more information). Porting it to Python allows for cross-platform usage.

Features

• Extracts the loudest section of a WAV file.
• Specify desired segment length in milliseconds.
• Skip audio files that are below a minimum volume threshold.
• Process multiple files using glob patterns.

Prerequisites

• Python 3 (tested on version 3.12.7).
• NumPy (tested on version 2.1.3).
• PySoundFile (tested on version 0.12.1).

All dependencies are listed in the requirements.txt file.

Installation

1. Clone the repository:

git clone https://github.com/flacle/extract_loudest_section_python.git
cd extract_loudest_section_python

2. Create and activate a virtual environment (optional but recommended):

python -m venv venv
source venv/bin/activate # On Windows use: venv\Scripts\activate.bat

3. Install the dependencies:

pip install -r requirements.txt

Usage

The script extract_loudest_section_python.py can be run from the command line. It requires at least two arguments: an input file path pattern (glob) and an output directory path.

python extract_loudest_section_python.py <input_glob> <output_directory>

Arguments

• <input_glob>: Glob pattern for input WAV files (e.g.,input/*.wav) that will match all WAV files in the data directory.
• <output_directory>: Path to the output directory where the extracted loudest sections (trimmed versions) will be saved.

Optional Parameters:

You can modify the desired length and minimum volume threshold by editing the variables desired_length_ms and min_volume in the main() function of extract_loudest_section_python.py.

Example

To process the WAV file in the sample_wav directory and save the trimmed version to the sample_trimmed directory, run the following command:

python extract_loudest_section_python.py sample_wav/*.wav sample_trimmed

Provided Sample WAV file

This repository contains one sample WAV file from the LibriSpeech dataset. The file was originally in flac, and was converted to 16-bit PCM WAV file with a sample rate of 16 kHz, using the following FFmpeg command:

ffmpeg -i 1688-142285-0000.flac -ar 16000 -acodec pcm_s16le 1688-142285-0000.wav

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Acknowledgements

• Pete Warden for the original C++ implementation of the extract_loudest_section tool. Original repository: extract_loudest_section.
• LibriSpeech dataset for the sample WAV file.

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.