https://github.com/dan-k-k/vocal-gate
Real-time AI Noise Gate VST3/AU plugin. Removes coughs, sneezes, and other artifacts from your live streams, podcasts, and videos.
https://github.com/dan-k-k/vocal-gate
ai-audio au-plugin audio-plugin content-creation cpp dsp juce machine-learning noise-reduction noise-suppression obs-studio onnx podcasting real-time-audio streaming video-editing voice-processing vst vst3 youtube
Last synced: 24 days ago
JSON representation
Real-time AI Noise Gate VST3/AU plugin. Removes coughs, sneezes, and other artifacts from your live streams, podcasts, and videos.
- Host: GitHub
- URL: https://github.com/dan-k-k/vocal-gate
- Owner: dan-k-k
- License: gpl-3.0
- Created: 2026-03-04T13:08:23.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2026-03-21T11:34:36.000Z (about 1 month ago)
- Last Synced: 2026-03-22T01:56:09.761Z (about 1 month ago)
- Topics: ai-audio, au-plugin, audio-plugin, content-creation, cpp, dsp, juce, machine-learning, noise-reduction, noise-suppression, obs-studio, onnx, podcasting, real-time-audio, streaming, video-editing, voice-processing, vst, vst3, youtube
- Language: C++
- Homepage:
- Size: 37 MB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## Vocal Gate VST3 & AU (Free) 
Vocal Gate is a free AI noise gate VST3/AU plugin trained to separate clean speech from unwanted microphone artifacts. This plugin is intended for Creators and Editors. Can be used as: a **cough remover** in your podcast/Youtube video editor, a **sneeze ducker** in your Twitch live stream broadcaster, a **filter** to block keyboard clicks and heavy breathing in your recording software, among other permutations.
Requires: macOS 11+ (M-series) | Windows 10+ (x64) | Linux (Ubuntu 22.04+ / x64) | Release notes
Note: Installers are unsigned. On macOS, right-click Open the installer in your downloads. On Windows, press 'More info' and 'Run anyway'. On Linux, extract the .zip and place the .vst3 folder into ~/.vst3/.
### Real-World Use Examples
Watch the full demo on YouTube
#### Useful for:
* **Live Streaming:** Twitch, YouTube Live, Kick, Facebook Gaming
* **Podcasting:** Spotify, Apple Podcasts, Patreon
* **Video Content:** YouTube VODs, TikTok, Instagram Reels
#### Works in:
* **Broadcasting:** OBS Studio, Streamlabs, vMix
* **Video Editing:** DaVinci Resolve, Adobe Premiere Pro, Final Cut Pro
* **DAWs & Audio:** Reaper, Logic Pro, FL Studio, Ableton Live, Audacity
> **IF RECORDING/STREAMING LIVE (OBS, Streamlabs, etc.):** The AI model requires a **750ms** lookahead to work. You MUST delay your visual sources and desktop audio by 750ms to keep everything in sync.
>
> **E.g., How to sync in OBS:**
> * **Webcams:** Right-click source -> *Filters* -> Add *Video Delay (Async)* -> set to 750ms.
> * **Game/Screen Capture:** Right-click source -> *Filters* -> Add *Render Delay*. (Note: OBS caps this at 500ms, so you must stack **two** Render Delay filters: one at 500ms and another at 250ms).
> * **Desktop Audio:** Click the gear icon in the Audio Mixer -> *Advanced Audio Properties* -> find your Desktop Audio and set the *Sync Offset* to 750ms.
---
## Model Performance
The plugin relies on a pruned and quantised int8 ONNX model to achieve real-time inference in just ~0.3 ms per buffer.
### ROC
The pruned and quantised model has better performance in both inference time and ability on the test set (it is better generalised).
### Dataset Energies
### Training Loss
