Real-time Noise Suppression Plugin (VST2, LV2, LADSPA)

A real-time noise suppression plugin for voice based on Xiph's RNNoise. More info about the base library.

Fork differences

Upstream rnnoise with AVX2 patch and additional models from GregoR
Configuration GUI using Dear ImGui
Make it STEREO (for good or bad, but sometimes you'll hear those independence)

About

The plugin is meant to suppress a wide range of noise origins (from original paper): computer fans, office, crowd, airplane, car, train, construction.

From my tests mild background noise is always suppressed, loud sounds, like clicking of mechanical keyboard, are suppressed while there is no voice however they are only reduced in volume when voice is present.

The plugin is made to work with 1 channel and/or 2 channels (ladspa plugin), 16 bit, 48000 Hz audio input. Other sample rates may work, or not...

How-to

Windows + Equalizer APO (VST2)

To check or change mic settings go to "Recording devices" -> "Recording" -> "Properties" of the target mic -> "Advanced".

To enable the plugin in Equalizer APO select "Plugins" -> "VST Plugin" and specify the plugin dll.

See detailed guide provided by @bssankaran.

Linux

PipeWire

You could use pulseeffects which is a nice GUI application with many useful filters, however I failed to make it work consistently to post-process the output of my mic, the below method should be more bulletproof.

See under "Example configuration of a virtual source" in the PipeWire documentation, you would only need to change plugin = ladspa/librnnoise_ladspa to plugin = actual/path/of/plugin/librnnoise_ladspa.so.

To make the change persistent you could add it to your pipewire.conf. According to the pipewire documentation the best practice would be to first copy the example pipewire.conf from /usr/share/pipewire/pipewire.conf into your own ~/.config/pipewire/pipewire.conf.

mkdir -p ~/.config/pipewire
cp /usr/share/pipewire/pipewire.conf ~/.config/pipewire/pipewire.conf

Now you can edit ~/.config/pipewire/pipewire.conf and under the context.modules section, add the Example configuration from the Pipewire docs to the end of the section. Paste it after { name = libpipewire-module-session-manager } to be more precise.

Save and restart the pipewire service.

PulseAudio

The idea is:

Create a sink from which apps will take audio later and which will be the end sink in the chain.
Load the plugin which outputs to already created sink (sink_master parameter) and has input sink (sink_name parameter, sink will be created).
Create loopback from microphone (source) to input sink of plugin (sink) with 1 channel.

For example, to create a new mono device with noise-reduced audio from your microphone, first, find your mic name using e.g.:

pactl list sources short

Then, create the new device using:

pacmd load-module module-null-sink sink_name=mic_denoised_out rate=48000
pacmd load-module module-ladspa-sink sink_name=mic_raw_in sink_master=mic_denoised_out label=noise_suppressor_mono plugin=/path/to/librnnoise_ladspa.so control=50
pacmd load-module module-loopback source=<your_mic_name> sink=mic_raw_in channels=1 source_dont_move=true sink_dont_move=true

This needs to be executed every time PulseAudio is launched. You can automate this by creating file in ~/.config/pulse/default.pa with the content:

.include /etc/pulse/default.pa

load-module module-null-sink sink_name=mic_denoised_out rate=48000
load-module module-ladspa-sink sink_name=mic_raw_in sink_master=mic_denoised_out label=noise_suppressor_mono plugin=/path/to/librnnoise_ladspa.so control=50
load-module module-loopback source=your_mic_name sink=mic_raw_in channels=1 source_dont_move=true sink_dont_move=true

set-default-source mic_denoised_out.monitor

If you have a stereo input use these options instead:

label=noise_suppressor_stereo
channels=2

If you have problems with audio crackling or high / periodically increasing latency, adding latency_msec=1 to the loopback might help:

load-module module-loopback source=your_mic_name sink=mic_raw_in channels=1 source_dont_move=true sink_dont_move=true latency_msec=1

⚠️ Chrome and other Chromium based browsers will ignore monitor devices and you will not be able to select the "Monitor of Null Output". To work around this, either use pavucontrol to assign the input to Chrome, or remap this device in PulseAudio to create a regular source:

pacmd load-module module-remap-source source_name=denoised master=mic_denoised_out.monitor channels=1

Additional notes:

You can get librnnoise_ladspa.so either by downloading latest release or by compiling it yourself.
You may still need to set correct input for application, this can be done in audio mixer panel (if you have one) in 'Recording' tab where you should set 'Monitor of Null Output' as source.

Plugin settings:

VAD Threshold (%) - if probability of sound being a voice is lower than this threshold - silence will be returned. By default VAD threshold is 50% which should work with any mic. For most mics higher threshold control=95 would be fine. Without the VAD some loud noises may still be a little bit audible when there is no voice.
There is also an implicit grace period of 200 milliseconds, meaning that after the last voice detection - output won't be silenced for 200 ms.

MacOS

You will need to compile library yourself following steps below.

It is reported that VST plugin works with Reaper after removing underscore from lib name.

Status

The plugin is tested with Equalizer APO v1.2 x64 (open source system-wide equalizer for Windows) and tested with pulse audio on arch linux.

I'm not associated with the original work in any way and have only superficial understanding of it. The original author will probably create something better out of their work but for now I don't see any analogs with similar capabilities so I have created a usable one.

Developing

VST sdk files aren't shipped here due to their license. You need to download VST sdk and copy several files to src/pluginterfaces/vst2.x/ and to src/vst2.x/. You can find sdk on archive.org since it was removed from the official website.

├── pluginterfaces
│  └── vst2.x
│     ├── aeffect.h              |
│     ├── aeffectx.h             | -> Copy into src/pluginterfaces/vst2.x/
│     └── vstfxstore.h           |
├── public.sdk
│  ├── samples
│  └── source
│     └── vst2.x
│        ├── aeffeditor.h        |
│        ├── audioeffect.cpp     |
│        ├── audioeffect.h       |
│        ├── audioeffectx.cpp    | -> Copy into src/vst2.x/
│        ├── audioeffectx.h      |
│        └── vstplugmain.cpp     |

LV2 and LADSPA sdk files are in repository.

All improvements are welcomed!

Compiling

For Windows you either need mingw or hope it works with visual studio cmake generator.

For MacOS steps are the same as for Linux.

If you did not download and place VST sdk files - VST plugin won't be built.

Compiling for x64:

cmake -Bbuild-x64 -H. -DCMAKE_BUILD_TYPE=Release
cd build-x64
make

Compiling for x32:

cmake -D CMAKE_CXX_FLAGS=-m32 -D CMAKE_C_FLAGS=-m32 -Bbuild-x32 -H. -DCMAKE_BUILD_TYPE=Release
cd build-x32
make

Cross-compiling for Windows x64:

cmake -Bbuild-mingw64 -H.  -DCMAKE_TOOLCHAIN_FILE=toolchains/toolchain-mingw64.cmake -DCMAKE_BUILD_TYPE=Release
cd build-mingw64
make

Turning plugins on and off

By default, all three plugins are being built if the necessary libraries are present. You can deliberately turn off plugins by using the following CMake flags:

BUILD_VST_PLUGIN
BUILD_LV2_PLUGIN
BUILD_LADSPA_PLUGIN

For example:

cmake -DBUILD_VST_PLUGIN=OFF

License

This project is licensed under the GNU General Public License v3.0 - see the LICENSE file for details.

shemanaev/noise-suppression-for-voice