DTLN a quantised C version would be great

[StuartIanNaylor]

Even having it as standalone as a LDSPA plugin would be great for Linux audio in general.
Deepfilter has a LADSPA filter that you can just use a Alsaplug to downsample but tract is single thread so it needs an overly large core to run well.
I did a simple beamformer https://github.com/StuartIanNaylor/2ch_delay_sum which is my very bad 1st hack of C/C++ and MVDR would be much better but couldn't work out if its casual or not or looks ahead in the source I had https://github.com/robin1001/beamforming and using a lib such as Eigen than the homebrew was beyond me, but reducing reverberation on input helps the filter and ASR.