Noise Gator vs RNNoise

Noise Gator processes microphone audio in real time to reduce noise, then sends the cleaned signal to a virtual audio device for other apps. It is free under the MIT license and runs on Windows and macOS. Discord, Teams, and Zoom are named as apps that can use the virtual device as a microphone input. DeepFilterNet is the default denoising engine, with dual-pass RNNoise as a lighter fallback. The audio pipeline also includes a highpass filter, suppression control, a voice-activity-driven noise gate, three-band EQ, and automatic gain normalization. Bluetooth headset audio is supported, and the device menu refreshes when devices connect or disconnect. The app can install VB-Cable on Windows or BlackHole on macOS automatically. Prebuilt binaries are available from GitHub Releases; the ZIP includes the executable and ONNX Runtime DLL. DeepFilterNet’s runtime and an approximately 8 MB model download on first use, with SHA256 verification. Settings are saved to a TOML file for the next launch, and headless mode supports scripted or server use. The latest listed release is v0.1.2. Building from source requires stable Rust 1.85 or later, plus MinGW-w64 or MSVC on Windows.

RNNoise has no write-up on Specifiction yet; its figures are above.