The CoreAudio framework has a process class property kAudioProcessPropertyDevices, which is used to obtain an array of AudioObjectIDs that represent the devices currently used by the process for output.
But this property behaves incorrectly. Specifically, if a process switches from one microphone to another while streaming, this property returns the output device ID as the input device ID.
Steps to reproduce:
- run FaceTime
- select "MacBook Pro Microphone" as an input device from the FaceTime menu
- select "MacBook Pro Speaker" as an output device from the FaceTime menu
- start a call
- get kAudioProcessPropertyDevices for Input scope: returns ID1 - "MacBook Pro Microphone" [CORRECT]
- get kAudioProcessPropertyDevices for Output scope: returns ID2 - "MacBook Pro Speaker" [CORRECT]
- change the input device in the FaceTime menu to any other microphone ("AirPods Pro" - ID3)
- get kAudioProcessPropertyDevices for Input scope: returns ID2 "MacBook Pro Speaker" but should be ID3 "AirPods Pro" [INCORRECT]
- get "kAudioProcessPropertyDevices" for Output scope: returns ID2 "MacBook Pro Speaker" [CORRECT]
Monitoring the property change for kAudioProcessPropertyDevices could provide a means to track audio streaming processes, but its current flaw renders it unusable.
So I'm curious if the macOS developers plan to address this issue in future releases, or if anyone can come up with a reliable alternative for identifying processes and associated audio devices being used for playback or recording.