Speaker Identification
Label who said what, entirely on your Mac.
Splits a transcript by voice and labels each segment, so a panel or interview reads like a script.
On by default. Turn it off in Settings ▸ General. It runs after transcription, so imports take a little longer with it on.
Naming Speakers
Speakers start as generic labels. In a video's speakers view, each speaker row has:
- Name speaker - give the voice a name, which relabels every segment attributed to it.
- Previous / next segment - step through that speaker's turns, useful for checking the app attributed them correctly before you commit to a name.
Once named, Silkwave Video stores a voice profile. When that voice appears in a later import it is recognized and suggested for you to confirm - useful for a podcast you follow or a colleague who recurs across videos.
Speaker Profiles
Settings ▸ Speaker Profiles lists everyone Silkwave Video has learned, and lets you delete a profile.
Deleting removes the profile and unlinks it from every video it was linked to; those speakers revert to unnamed. This cannot be undone.
Privacy
Profiles are created and matched on your Mac, never uploaded, and never sent to the AI features - those receive transcript text only.
Accuracy
Matching works from voice characteristics, so it degrades with overlapping speech, similar voices, background music over speech, or a speaker who appears only briefly.