It seems like it's programmatic. Head-related transfer functions, or something starting with those letters.
But they're not written out in code somewhere. They're all built in to inaccessible commercial programs.
And I'd be more interested in applying it to games, anyway, so even being able to create files that use 3d sound wouldn't really work so well.

And elevation perception apparently has a lot to do with the shape of one's ears and such, so it's possible for people to perceive artificial 3d sound differently just based on having differently-shaped ears...

The annoying thing isn't that it's complicated, so much as that people can and do use it. A very select few people. Anyone that can use the internet has access to programming instructions, basic audio-editting software, Etc. But getting a hold of this stuff is like winning the lottery.

... There's a difference?

But... yeah... both would be nice.

Spectral peaks and notches, of course! Whatever those are / however those are edited.
The shape of the outer ear applies its own filters, which will vary based on the direction of the sound. The brain does the rest. The easiest way to simulate this in a recording would be with binaural (SP?) microphones stuck in either actual ears, or a dummy head with realistic ears. The poor quality thing I've done by flipping my headphones inside out can sometimes pick up vertical positioning, especially if I've got something behind each ear, but that doesn't do me much good for applying it to sounds that already exist.

As for the horizontal positioning, what I've read suggests our brains use both methods, with one being suited better for low frequency sounds and the other for sounds in the range of human speech. It's actually a little amusing that it sounds like we have three different mechanisms for determining position, but two of those are devoted to the one effect everyone knows how to do already.