The user suggests that the fast-vad tool could benefit from multi-threaded processing to improve performance, especially for real-time applications.
Repo: https://github.com/AtharvBhat/fast-vad I needed something comparable to existing open-source VADs in quality, but with a strong emphasis on speed, simple integration, and streaming support. To my knowledge it's the fastest open-source VAD out there. Highlights: - Rust crate + Python package - batch and streaming/stateful APIs - built-in modes for sensible defaults - configurable lower-level knobs if you want to tune behavior yourself It's a simple logistic regression that operates on frame based features to keep it as fast as possible. It was trained using libriVAD dataset ( small version ) If anyone works on Audio, do try it out and let me know how it goes ! Feedback would be helpful 🙂