audiblelight 0.1.0.post4

A library for soundscape synthesis using spatial impulse responses derived from ray-traced room scans

AudibleLight

A Controllable, End-to-End API for Soundscape Synthesis Across Ray-Traced & Real-World Measured Acoustics

⚠️ WARNING: This project is currently under heavy development. We have done our due diligence to ensure that it works as expected. However, if you encounter any errors, please open an issue and let us know.

Contents

• Installation
• Usage
• Contributions
• Roadmap

What is AudibleLight?

AudibleLight is a unified API for soundscape synthesis supporting ray-traced, real-world, and parametric RIR generation. It enables flexible microphone array modeling and dynamic, fully annotated source trajectories within a single workflow. It is built upon SpatialScaper, SoundSpaces and Pyroomacoustics for scalable soundscape generation with unprecedented acoustic diversity.

AudibleLight is developed by researchers at the Centre for Digital Music, Queen Mary University of London in collaboration with Meta Reality Labs.

Installation:

Prerequisites

• python3.10 or above (tested up to python3.12)
• A modern Linux distro: current versions of Ubuntu and Red Hat have been tested and confirmed to work.
  • Using another OS? Let us know so we can add it here!

If you're looking to develop AudibleLight, you'll also need:

• git
• poetry
• make

Install via pypi

For non-development installs, the simplest way to install AudibleLight is via pypi:

sudo apt install -y libsox-dev libsox-fmt-all freeglut3-dev pandoc
pip install audiblelight

Install via the command line

git clone https://github.com/AudibleLight/AudibleLight.git
cd AudibleLight
make install

Download datasets

We provide several helper scripts to download and prepare data (3D meshes, sofa files, audio files) that may be useful in AudibleLight. To run these:

make download

For further information, see scripts/download_data/README.md.

Usage

See the quickstart guide for help getting started with dataset generation in AudibleLight

Script

We include scripts to generate large datasets for common tasks relating to spatial soundscape synthesis.

To generate a dataset for sound event localization and detection (SELD) that conforms with the DCASE2023 task 3 format, run:

poetry run python scripts/seld/generate_dataset.py

To see the available arguments that this script takes, add the --help argument

If you want to generate custom datasets (or want to implement AudibleLight in a larger pipeline), please see the quickstart, tutorials, and API documentation.

Contributions

... are welcome! Please make a PR or take a look at the open issues.

Running the tests

Before making a PR, ensure that you run the pre-commit hooks and tests:

make fix
make tests

Roadmap

• Spatial audio augmentations (from https://arxiv.org/abs/2101.02919)
• HRTF support
• Directional microphone capsules support
• Increased visualisation options