StuartIanNaylor

`--features=agc` still much larger 33mb+ vs 24mb+ of the release, does it need stripping? The plugin runs is there now a control level for the AGC as struggling to notice...

Yeah --profile=release-lto solves that and also the level seems to be amplitude not RMS which is also great as RMS is harder to visualise. I am not sure as think...

@Rikorose Hendrik there is a problem somewhere with the AGC. If you have high noise say play some music on speaker close by then with no AGC the filter works...

The voice that should be there is gone apart from fragments, you need to up the the SNR high and with AGC for some reason the voice will start to...

Lows low SNR I mean or high db noise https://imgur.com/gallery/qnjGwLP Top agc off bottom agc on ``` pcm.ladspadf { type ladspa slave.pcm "plughw:3,0"; path "/usr/lib/ladspa"; plugins [{ label deep_filter_mono input...

logs ``` 2022-12-15T19:55:05.564Z | INFO | df::tract | Init DF decoder with delay: 0 2022-12-15T19:55:05.573Z | INFO | df::tract | Running with model type deepfilternet3 lookahead 2 2022-12-15T19:55:05.573Z | INFO...

Compiled from the agc branch earlier

``` git status On branch feat_agc Your branch is up-to-date with 'origin/feat_agc' ``` ``` RUST_LOG=TRACE arecord -Ddeepfilter -r16000 -fS16_LE -c1 test2.wav ``` I will compile again to check but reasonably...

Apols Hendrick I thought there was just 2 controls as in ALSA its just ordinals so thought it was the 1st 2 It should be controls [ 84 xx 0.5...

controls [ 84 -10 30 20 0.2 ] just going to try as seem to be defaults apart from att & agc