Speaker embedding
A speaker embedding is a compact numeric vector representing the characteristics of a voice, such that the same speaker's clips land close together — the machinery behind diarization and speaker verification.
Embeddings turn 'same voice?' into a distance measurement. Diarization clusters them; verification compares them to an enrolled profile; both inherit their biases from the embedding model's training data.
Related terms
Frequently asked
What is Speaker embedding?
A speaker embedding is a compact numeric vector representing the characteristics of a voice, such that the same speaker's clips land close together — the machinery behind diarization and speaker verification.
Why does Speaker embedding matter?
Embeddings turn 'same voice?' into a distance measurement. Diarization clusters them; verification compares them to an enrolled profile; both inherit their biases from the embedding model's training data.
What terms are related to Speaker embedding?
Closely related concepts: Speaker diarization, Speaker identification, Voice cloning — each has its own entry in this glossary.