I'm trying to send live audio to a neural network in PyTorch which will in turn send OSC messages to SuperCollider, but currently the PyTorch end of things isn't listening to Jack, on which SuperCollider is dependent. The NN in PyTorch receives audio just fine on default PulseAudio via a USB interface, but as soon as I add Jack to the chain it stops recognising audio in.
I'm aware that the fix might well involve digging into the settings on the NN side, but a very proficient Python programmer checked it out briefly yesterday and didn't find a solution, and if there was a way of either making Jack present the audio input as if it was via PulseAudio, or of running SuperCollider without PulseAudio, it would be problem solved.
Thanks!
Mark