Skip to content

Fully open-source and state-of-the-art Voice Activity Detection (VAD) models for academic research and commercial applications.

License

Notifications You must be signed in to change notification settings

stefanwebb/open-voice-activity-detection

Repository files navigation

Fully Open-Source Voice Activity Detection (VAD) for Real-Time Speech Applications

Voice Activity Detection (VAD) is a critical first step in any application involving speech recognition. However, while exploring real-time voice chat agents, I found that many state-of-the-art (SoTA) models are not truly open-source—they provide only open weights, limiting transparency and hindering research and development.

This repository aims to change that by providing a fully open and research-friendly implementation of the Silero VAD model. The goal is to advance the state of the art in VAD through open experimentation, training, and integration.

Status

As of May 27, 2025, this repository includes:

✅ A complete implementation of the Silero VAD model for research use

Roadmap

In the near future, I plan to add the following:

🧠 Code to train Silero VAD from scratch on custom datasets

📊 Evaluation scripts for standard VAD benchmarks

🔧 Support for LoRA fine-tuning to extend or adapt Silero VAD

🔌 Example integrations with Python, client-side web applications, and Unity

Instructions

Install the package in editable mode:

pip install --editable .

License

This project is released under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0), encouraging both academic research and commercial application.

About

Fully open-source and state-of-the-art Voice Activity Detection (VAD) models for academic research and commercial applications.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published