What speaker labels can tell you
Diarization marks portions of speech as belonging to different speakers. It helps you follow the questions and answers in a longer conversation. The model separates voices; you still need to match each label to the actual participant.
Overlapping speech or interruptions can produce an incorrect label. Check those boundaries before using the text for an interview draft or a meeting recap. A clear label is useful only when its attribution is right.
Prepare both recognition and speaker models
- Open Settings → Transcription in Whynote on Mac.
- Download Local Model (FluidAudio) if it is not ready.
- Under Speaker Diarization, download the additional speaker model.
- Turn on Enable Speaker Identification once the download is complete.
- Select local note transcription for the workflow you intend to use.
The recognition model supplies the words, while the additional model separates the voices. A finished transcript without labels does not confirm that both parts are ready. Prepare them before disconnecting with the offline setup guide.
Set an expected speaker count when you know it
Use Expected Speakers in the diarization settings to give the model a known participant count. For a two-person interview, choose two; when you do not know the count, leave automatic detection selected. The setting is a hint about the conversation rather than a list of people’s names.
Start with a short passage where each person speaks separately and review the labels. If similar voices are grouped together or one voice receives several labels, compare those passages with the recording. A preset count still needs that check.
Check attribution before writing the outcome
Listen to the first lines from each participant and match them to the transcript labels. Review interruptions, brief replies and passages recorded after a microphone change. Leave an uncertain attribution open until you can confirm it.
A speaker label should not become proof of who accepted an action. Check the person and wording against the recording before adding the task to your notes. The label is an aid for finding the passage you need.
When labels are missing or inaccurate
| Situation | What to check |
|---|---|
| Text appears without labels | Confirm the separate speaker model is downloaded and identification is enabled. |
| The control is unavailable | Wait for the diarization model to be ready rather than only starting its download. |
| One person receives several labels | Listen to the passages and check the expected speaker count. |
| Two people share a label | Check overlap and audio quality, then correct important attribution manually. |
If speaker identification cannot complete, the text may still be available without labels. Keep the recording so you can check important attribution yourself. The privacy guide explains local processing separately from account synchronization.
Bring your next recording into Whynote
Get Whynote for Mac to record a conversation or open a saved audio file. Choose the iPhone app for voice notes and recordings you keep on your phone.