Yes, I specifically avoided mentioning that because they don't need to diarize from audio directly as they know exactly which participant is speaking at all times.

They have access to each individual’s stream in that case in order to diarize.