Capturing any kind of audio can be done using any number of microphones and audio engineering techniques. Ideally, you’ll want to record each element of a performance–vocals, the instruments, and so on–on separate tracks so you can mix everything together afterward, a process more complicated but in the end more convenient and precise than just using one microphone to record everything together. This new tech, however, takes things in a completely different direction.
Researchers at Carnegie Mellon University figured out how to use a couple of ordinary video cameras. But turning them into “optical microphones,” the cameras sense the vibrations and frequencies of the individual elements of a performance and hone in on them. There are lasers and complicated algorithms involved that are best explained by this video.
A deeper dive can be found here.