As I have been doing more videos and streaming I have been working on a noise suppression filter for the human voice. Currently I have written this as a LV2 plugin, however I am in the process of porting it into an OBS audio filter.
This is NOT a simple gate/expander/compressor setup, nor does it identify voice by differentiating the stereo channels. The input is first downmixed to mono before processing.
This plugin only incurs approximately 30ms of latency which is low enough for real time monitoring without speech jamming.
Right now I am currently live re-streaming a friend (bamfsvamps) via the voice plugin.
Edit: Finished, but you can see the recording here:
Note the filter is extremely effective at removing even the game audio which is mixed into the output.
Edit: Also @bamfsvamps is using a laptop to play and is recording via the laptop mic, so the filter is working very hard to remove her keyboard noise. When using a more conventional setup (cheap headset w/mic) the key nose is entirely suppressed.
Note: This is NOT an open source project and will be for sale when it is ready, price is TBD but I am aiming to keep it affordable for those with a limited budget and are just streaming for fun.