Speaker embeddings

Also written: speaker embedding

A vector from a neural network that is similar for the same voice. Diarization clusters these vectors to decide who is who.