GlossaryQuality
Speaker labels
Also known as: speaker identification
Labels in a transcript indicating who is speaking at each segment — often displayed as 'Speaker 1:' or replaced with actual names in post-processing.
Speaker labels make transcripts of multi-speaker events readable. Without them, a panel discussion appears as undifferentiated text where the reader can't tell who said what.
Speaker labels are produced by speaker diarization (acoustic analysis identifying voice changes). The output is typically generic labels like 'Speaker 1' that can be mapped to actual names in the dashboard once the event ends.
For events with crisp per-speaker microphones, diarization quality is high. For events where audio is mixed at a single room mic, diarization is noisier and may require manual correction.