Just as each person has a unique set of characteristics that make up their faces, our voices also have a wide array of measurable characteristics that uniquely identify who we are.

Our voiceprints are what enable smart Bluetooth speakers like Amazon’s Alexa to recognize different people within the same space. New machine learning techniques, combined with vast datasets of recorded voices, have now enabled researchers to identify us simply by listening to the micro-signatures produced when we speak.

Researchers at Carnegie Mellon University discovered a generative technique allowing them to build a 3D version of someone’s face using only their voiceprint.

This system is being deployed by law enforcement agencies to identify prank callers in a practice known as “swatting,” in which they make false claims to local agencies, which then unwittingly deploy emergency crew to false emergencies.