Is it possible to detect the Window that is making an app?

Making a sound*, not making an app.

Is it possible to detect the Window that is making a sound?

The short answer is no, not really. First off, from the system perspective, processes play audio, not windows. Obviously many apps do present a connection between a window and a given audio stream, but that's something the app is imposing on it's own content not something the system was actually "aware" of.

Making that concrete, Safari will let you mute/unmute individual windows but what Safari is actually doing is "manually" muting the individual streams* it's feeding into the system. The system has no idea that there's ANY connection between those streams and specific windows.

*Assuming it's even giving the system multiple streams. It's an app can also be doing it's own mixing, then passing a single audio stream "out" to the system.

So, changing the question "is it possible to detect which processes are producing audio" the answer is yes... but not really. More specifically:

1. CoreAudio's "kAudioDevicePropertyHogMode" will let you detect what (if any) process has taken exclusive control of the audio interface. That's straightforward to use, however, I don't think it'll be all that useful. Seizing the interface like this is relatively rare and the kind of apps that do it (basically "games" and some media players) aren't the kind I'd typically think of trying to capture.

2. If you look at the broader app ecosystem, you'll find that there are apps that clearly "know" which apps are playing audio, as they're able to do things like alter the volume or mute individual processes. These apps actually work by installing a system wide audio driver which lets them act as the "target" output device that other apps are then sending audio to. They could easily be bypassed either by the user (who could change to a different output device) or individual app (if the app chooses to take control of it's own routing), but in practice that's not really a problem since the user "wants" to be using them and mosts apps just use the default route.

The second approach is certainly workable (as many apps obviously do this), but the work involved is... quite significant. Most of the apps that do this sort of thing are in fact "dedicated audio apps" as the work involved in making something like this "from scratch" basically requires that kind of effort and expertise.

That leads to the other option here, which would be to use someone else's code/library. Some commercial developers also license their routing engines and there are some GPL projects as well.

__
Kevin Elliott
DTS Engineer, CoreOS/Hardware