This project runs spleeter as a pulseaudio module to filter all the sound going out of the computer.
git clone https://github.com/mie00/spleeter-realtime
cd spleeter-realtime
# install pulseaudio plugin
cd pulseaudio
make
sudo make install
cd ..
# install python dependencies
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
# test if tensorflow works on gpu
python test_gpu.py
./run.sh
press q or ctrl+c to quit, any key to toggle it on/off.
- Run the server
cd spleeter-realtime
source .venv/bin/activate
python spleeter_realtime.py
- Get the master sink
~ pactl list sinks short
0 alsa_output.pci-0000_00_1f.3.analog-stereo module-alsa-card.c s16le 2ch 44100Hz RUNNING
~ export MASTER=alsa_output.pci-0000_00_1f.3.analog-stereo
In this case alsa_output.pci-0000_00_1f.3.analog-stereo
is the sink name that is going to be used.
- Load the plugin
~ pactl load-module module-ladspa-sink plugin=spleeter_ladspa label=vocals sink_master=$MASTER
26
26 is the module id used to unload the plugin run pactl unload-module 26
to unload it.
- Set default sink to the
LADSPA Plugin Vocals on ...
sink, using eitherpavucontrol
,pactl set-default-sink
or any other sound control software.