If you send your apps' audio to any device, physical or virtual, the OS itself provides a loopback for that, which is what the Audio Output Capture picks up. No plugin for that at all.
If you send your apps' audio to VB Cable, which I also see in that list, it provides its own loopback in the form a virtual mic, that an Audio Input Capture can pick up. Still no plugin at all, unless you're calling VB Cable a "plugin".
From googling
blackhole sound mac
, I got what looks like their homepage and GitHub:
Zero Latency. Perfect for Streamers, Podcasters, and Online Instructors.
existential.audio
BlackHole is a modern macOS audio loopback driver that allows applications to pass audio to other applications with zero additional latency. - ExistentialAudio/BlackHole
github.com
And I don't see a difference at all between it and VB Cable, except that Blackhole supports 2, 16, or 64 channels, compared to 2 only with VB Cable.
One possibility is that two instances of Blackhole are not supposed to exist simultaneously? So I wonder if your 2-ch and 16-ch instances are trying to share something that shouldn't be shared, and getting confused between themselves?
The log has nothing about that, but I did notice:
14:59:20.715: [Loaded global audio device]: 'Mic/Aux'
14:59:20.715: - filter: 'Compressor' (compressor_filter)
14:59:20.715: - filter: 'Noise Gate' (noise_gate_filter)
14:59:20.715: - filter: 'Noise Suppression' (noise_suppress_filter_v2)
14:59:20.715: - filter: 'Gain' (gain_filter)
Is that the actual order, or is it reversed? As a programmer, I can see it either way. As an audio guy, the order shown seems exactly backwards to me.
If the Noise Suppression does indeed follow the Compressor and Gate, then its sense of noise is going to be all over the place, and it will let some through as the noise level changes. If the Noise Suppressor is first, or following only a fixed Gain and/or EQ, so that nothing changes, then its constant sense of noise allows it to remove more without killing your voice. Then you can do your dynamics on the de-noised version.