We have support for VOIP in Steam only at present, with the intention of coming up with a more cross platform implementation in the near future, possibly using the Opus codec.
Steam does not support “positional” voice audio per se, it just polls the microphone and compresses the data. We transport that data over the wire, uncompress it and play it back with an audio component.
At present that audio component is just marked as UI sound and doesn’t really have anything positional about it. It might be possible to extend that in the future to attach the audio component to a pawn/object in the world such that it would have 3d characteristics. I’d have to investigate that further and its not currently on my roadmap.