Glossary

What is speaker diarization?

Diarization is a machine working out who is speaking when, from the audio alone, and splitting the transcript by speaker automatically. Cast does not do this yet — you label speakers yourself, which takes a few clicks and stays right because you are the one doing it. Automatic detection is on the roadmap.

Two microphones give two separable tracks and nothing has to be guessed. One shared microphone gives a single blended waveform that no editor can reliably pull apart — which is why Cast asks you to label speakers rather than guess. — One mic each is separable. One shared mic is not — whatever the marketing says.

It is worth knowing the word because tools advertise it, and worth knowing its limits: diarization degrades badly on people talking over each other and on two voices captured by one microphone, which is exactly when you most want it to work.

Automatic speaker detection is on the Cast roadmap — see the Coming soon page to vote for it. Your own labels would keep priority over anything the model guesses.

Last updated 14 September 2026