There are a couple methods to detect "liveness" including depth perception of stereo-capable cameras, measuring reflection of IR emitter, eyeball movement, etc.
Source: work at an FR startup
Does it require specialized equipment? If yes, what’s the value of video vs fingerprint reader that’s more ubiquitous?
I don't think it was foolish. Its definitely going to be a challenge to identify fake videos in the near future
Perhaps 2FA through facial recognition + voice authentication?