A noise suppression CLAP / LADSPA / LV2 / VST2 / VST3 audio plugin based on a recurrent neural network, specially tailored for speech and voice.
The heavy lifting is done by the RNNoise project, this plugin is mostly a wrapper around that so that we can use it in real-time and within a regular audio plugin host.
This plugin has a fixed latency of 10ms, as that is the processing block size from RNNoise.
Also, THIS IS A WORK IN PROGRESS.
Progress so far:
- Automated build for Linux, macOS and Windows
- Optimized builds under x86 or NEON
- Exported shared library through MAPI (to be used for web-assembly)
- Report latency to host
- Click-free bypass
- Add "VAD Threshold" alike in werman/noise-suppression-for-voice
- Smooth mute/unmute transition for "VAD Threshold"
- Add custom UI
- Dynamic resampling (RNNoise expects 48kHz rate)
- More comprehensive UI
- More comprehensive README and documentation
- Make a 1.0 release
- Package it in KXStudio repositories
Because people will ask for it, current screenshot:
