Every now and then, it’s worth diving into the copious amounts of AR/VR research that Meta Platforms publishes online. Not all of it has clear ties to the company’s product ambitions, but a recently published paper reveals how Reality Labs researchers have progressed toward giving AR glasses users what Chief Scientist Michael Abrash has pitched as “audio superpowers.”

In particular, he’s talking about ways AR glasses could eventually deliver  noise-canceling and hearing assistance features. In order to make those kinds of “superpower” abilities work in a variety of situations, the glasses will need to be able to pinpoint the positions of people who are speaking. That’s not as simple as it might seem. Imagine you’re in a crowded room: figuring out who’s talking isn't always straightforward. Meta’s new paper, submitted to next month’s Conference on Computer Vision and Pattern Recognition, details a new approach to that problem, known as “active speaker localization.”