Improve Speech from Noise

[angrave]

e.g. Some old and new ideas mentioned here (check response dates)
https://superuser.com/questions/733061/reduce-background-noise-and-optimize-the-speech-from-an-audio-clip-using-ffmpeg

Some examples
"highpass=200,lowpass=3000,afftdn"
or maybe
highpass=f=200,lowpass=f=3000,arnndn=m=cb.rnnn
(see notes on getting the rnnn)

See also
https://github.com/xiph/rnnoise
https://github.com/GregorR/rnnoise-models