Put names to the voices in a meeting
Recordstead separates a recording into voices, but it cannot know whose they are. It asks you once per meeting, and the names you give go into that meeting's transcript and its summary. If you have saved voices and it recognizes every one of them, it does not ask at all.
Open the window
Section titled “Open the window”Nothing opens by itself. When the speakers step finishes with a voice it could not put a name to, a Review Speakers notification arrives, and the meeting waits for you: its row in Meetings reads Needs speakers, and it is listed on Home among the meetings that need you.
Open the window any of these ways:
- Click the notification.
- Choose Recording ▸ Review Speakers… (⌘N).
- Click the Review Speakers button on the meeting's row on Home.
- Open the meeting and click Review speakers on its Speakers step, which reads Waiting for you to name speakers.
The rest of the meeting is already done by then: the transcript is written and the voices are separated. Only the names are missing, and the summary is written once you answer.
Put a name to each voice
Section titled “Put a name to each voice”The window lists one row per voice, numbered in the order they first speak, with how long each one talked. The numbering is Recordstead's placeholder, not a guess about who they are.
- Play a sample. Each row offers a few, cut from that voice alone, so you can hear who it is rather than deduce it from the transcript.
- Type into the row's Name field. Once you have named people before, their names appear as chips under the field, and clicking one fills it in.
- Leave a row empty if you do not know. It is saved as an unidentified speaker and the transcript keeps its number.
When two rows are the same person
Section titled “When two rows are the same person”Recordstead says so itself when two voices measure as close: the row offers Same Person, which merges them, and Different People, which dismisses the question. It never merges anything by itself, because an extra voice you can merge costs less than two people wrongly joined.
When one row is two people
Section titled “When one row is two people”A row whose samples do not sound alike offers Samples Are from Different People…, which opens Separate this speaker. It says what to do: Play each sample, then choose who is speaking.
Mark each sample Speaker A or Speaker B, or Not sure for one you cannot place, then press Separate into Two Speakers. Both A and B have to be used at least once before that button works. The link appears only on a row with more than one sample, since a single sample is nothing to compare.
Save, or keep it for later
Section titled “Save, or keep it for later”Save Names writes the names into the transcript and lets the meeting finish. Keep for Later leaves the question open and the rest of the queue moving; the window reopens from any of the three doors above.
An unanswered review is not kept forever. After seven days Recordstead drops the question, and the meeting keeps the numbering it already has.
Change the names on a finished meeting
Section titled “Change the names on a finished meeting”Open the meeting, then choose Edit speakers… from its actions menu. The same window opens under the heading Edit speakers, and its buttons read differently there: Save changes to keep the edit.
Result
Section titled “Result”The meeting's transcript and summary print the names instead of the numbering, everywhere that voice speaks.
When it does not ask
Section titled “When it does not ask”If every voice in a meeting is one Recordstead has saved, and no two of them sound close enough to be the same person, the question never comes up: the names are written, the summary follows, and no notification arrives. The meeting's Speakers step says where the names came from, reading Named from your saved voices.
This can only happen when you have turned on recognizing saved voices, which is off until you do. One person it has never heard before, or two voices it cannot tell apart, brings the whole meeting back to you.
Nothing is locked by it. Edit speakers… opens the same window on any finished meeting, with the same samples to play.
What naming does not reach
Section titled “What naming does not reach”The names belong to that meeting. Recognizing the same voice in later recordings is a separate setting that is off until you turn it on, and the people it saves are kept per workspace.
A meeting whose recording has been removed has no samples left to play. You can still rename its speakers; you just cannot listen to them first.
A summary written before you renamed a speaker keeps the old name until you ask for it again, the same way a corrected transcript does: Update the summary afterwards.