Attach speakers

Attach times and speakers, such as [00:34] Speaker 2, to a transcribed meeting session. Sogon doesn't transcribe the text again, so this also works on past sessions.

There are two ways:

Method	Requirements	Does audio leave your Mac
On-device speaker attribution	macOS 26 or later, session audio	No
AWS Transcribe	The aws CLI, an AWS account, turned on in Settings	Yes, uploaded to S3

Attaching on device

1. Expand the session in Settings → History → Meetings.
2. If five or more people attended, enter the number in Attendees. Leave it blank for four or fewer.
3. Press Attach speakers. The first time, an approximately 234MB diarization model is downloaded.
4. Choose Speakers in the selector to see the result.

If a result already exists when you enter a count, press Redo to apply it.

Meetings with five or more people

With the count blank, the app separates at most four speakers. If five or more attended and the count is blank, the speech of the missing people gets distributed among the others. When the result has exactly four speakers the screen says so; enter the count and press Redo.

Naming speakers

Enter a name for each speaker in the Speakers tab, and the transcript and .txt/.md exports use it. Clear it to go back to the number. Names apply only to that session; Speaker 1 in the next meeting may be someone else.

Good to know

• In sessions that recorded both the microphone and system audio, Sogon treats the microphone side as one person, you, and splits only the system audio side into speakers.
• Even the same speaker gets a new timestamped line after a pause of more than 2 seconds.
• A time-match rate is shown next to the result. The lower it is, the more the times rely on estimation.
• Sogon removes speakers who spoke less than 2% of the total.
• Speakers don't appear in the minutes or verbatim outputs, because summarizing easily mixes up who said what. Export from the Speakers tab when you need speakers.