Speaker identification

When a recording stops, Minutist works out who spoke each line and labels them Speaker A, Speaker B, and so on. It does this by clustering the voices in the recording (diarisation), on-device. A rougher pass runs live while you record; the pass after you stop is the authoritative one and may relabel.

Turn it on or off under Settings → Processing → Identify speakers (on by default).

Naming speakers

On a saved meeting, click a speaker chip in the transcript to rename it; Speaker A becomes Alex, say. The name then appears on every line that speaker spoke. Renaming is an overlay: the transcript keeps the underlying letter, so the name is presentation, not a rewrite of the record.

Speaker chips are editable only on a finished, saved meeting. During a live recording the labels are provisional and get re-lettered when the meeting finalises, so they are shown but not yet editable.

Recognising a voice across meetings

Once you have named a voice, Minutist remembers it as a voiceprint and can recognise it in later meetings. When it is fairly but not certainly sure, it asks in the transcript (Is Speaker A Alex?) with Yes and No. Yes applies the name; No tells it that match was wrong.

Manage enrolled voices in the Voiceprints section of Settings. There you can rename a voice, merge two entries that are the same person, or forget one.

Note

Voiceprints stay on your machine. They are acoustic centroids used to label speakers; they are never uploaded.

Overlapping speech

When a segment overlaps more than one voice, Minutist marks it (for example, 2 speakers) rather than splitting it. The count tells you the line was shared.

Talk time

Ask the chat agent for per-speaker talk time and it reports how long each speaker held the floor, using the names you have set.

Caution

Reprocessing a recording re-runs diarisation from scratch, assigns fresh letters, and clears the names you set on that meeting.