How to Use Speaker Diarization in MacWhisper

Before you start

Speaker diarization adds speaker labels to a transcript so you can tell which words belong to which person. In MacWhisper, this works with supported transcription models after the audio has been recognized. The processing runs locally on the Mac by default, so the recording does not need to be uploaded to a service. Verify that your version of MacWhisper includes the diarization option before relying on it for a project.

Use a recording with clearly separated voices and limited background noise. A cleaner input usually produces more stable speaker labels. Check that the input device or file was recorded at a consistent input level and sample rate, because uneven audio can make speaker changes harder to detect.

Step-by-step instructions

• 1. Open MacWhisper on your Mac.
• 2. Add the audio or video file you want to transcribe.
• 3. Select a transcription model that supports speaker detection in your version of the app.
• 4. Enable the speaker diarization or speaker separation option if it is shown in the transcription settings.
• 5. Start the transcription and wait for the file to finish processing.
• 6. Review the transcript for speaker labels such as Speaker 1 and Speaker 2.

The exact menu path may differ between versions. If the option is not visible, check whether the selected model supports it or whether a newer version of MacWhisper is required. You can rename generic speaker labels after processing if the app provides that option.

Check the output

Play back short sections where the speaker changes and compare the labels with the voices you hear. This is more reliable than judging the transcript only by how smooth it looks. Look for places where one person is split into two labels or where two people share one label.

Diarization labels identify speaker turns, not real names. A label such as Speaker 1 is not stable across separate transcripts unless the app supports assigning or reusing speaker names. Expect more errors during overlapping speech, crosstalk, quiet voices or significant background noise.

Common problems and next steps

If speakers are merged, try a section with less overlap and stronger voice separation. If one speaker is split into multiple labels, check for changes in volume, distance from the microphone or background noise during the recording. Improving the input often helps more than transcribing the same file repeatedly.

When the labels are wrong in a few places, correct them manually if the editor supports renaming labels. For a recording that must be handled privately, confirm where MacWhisper processes files and whether any optional services are enabled before using it with sensitive audio.