Just to be clear, what are your end goal?
Split output to different outputs (headset vs. speakers) or just exclude/include certain sound sources from the stream? Combination of both?
EDIT: To be able to do the following, you either need Voicemeeter Banana or an additional Cable;
http://vb-audio.pagesperso-orange.fr/Cable/index.htm. This due to the limitation of only one virtual input with regular Voicemeeter.
If you've done the proper prerequisites correctly (configured your default Windows playback device, specified music player output etc) take a look at the following:
To split output you just need to configure A1 to be your headphones and A2 to be your speakers and route your inputs according to your liking.
B1 is what will be audible on stream, so if you don't want you music in your headphone or speakers, make sure only to send that sound to B1
Take a look at the screenshot I linked to above. There you can see my microphone only sending to B1. If I had A1 active I would hear the mic in my headphones. As you can see I have another channel called Spotify where I at the time of the screenshot output my music both to A1 and B1, i.e. both headphones and stream. Normally this would only be set to B1, not A1.
The first Virtual Input is where I input my default Windows playback device. It's where all normal sounds will end up, like games, web browser etc etc.I send this audio both to my headphones (A1) and the stream (B1).
Unused input channels are the Hardware Input 2 and the virtual input Voicemeeter AUX.
Unused output channels are A2, A3 and B2.
No matter if you use VAC or Voicemeeter there will be some stuff to figure out. First is to get how all the things are connected, the second and biggest part is to accept how Windows needs to be configured. The way of thinking is pretty much identical no matter if you choose VAC or Voicemeeter.
I like Voicemeeter better due to that you actuallt get a (very good) mixer a really nice GUI. All the FX and extras are just that ... extras :)