audio input channel handling
Currently, audio input handling assumes a mono stream only. If a multi-channel device is used, either the interleaved samples will be interpreted as a mono stream and/or buffer sizes will be incorrect. In either case, bad audio will be fed into the AI model.
The audio handling should be updated to allow for setting the desired channel and pulling only those channel's samples out of the interleaved stream. The inout channel should also be settable at run time as a commandline option.