Who said what: speaker labels
Turn on speaker separation, rename the speakers, and know when it struggles.
Speaker labels split the transcript by voice, so an interview reads as a conversation instead of one long block. It is available on Standard and Premium — Free plans get the text without the labels.
Turning it on
- 1Enable speaker labels in the upload options before you start, or run it on a finished transcript from the transcript view.
- 2If you know how many people are talking, say so. A definite number is a strong hint and it measurably improves the result.
- 3Wait a little longer than usual — this is a second pass over the audio.
Afterwards you get Speaker 1, Speaker 2 and so on. Rename them once and the new name replaces every occurrence in the transcript and in anything you download.
When it struggles
- People talking over each other — the hardest case there is.
- Two voices with a similar pitch and accent, especially on a phone recording.
- Someone who says three words in an hour. Very short turns often get folded into a neighbour.
- One shared microphone in a large room, where everyone sounds equally distant.
Important
Speaker labels are a best guess, not a legal record. If the transcript is going somewhere that matters, read it against the audio before you rely on the attribution.
Clean audio helps here more than anywhere else — how to get a cleaner transcript applies directly. Labels also carry into subtitles and the DOCX download.
Tags
Related Articles
How to get a cleaner transcript
The handful of things that genuinely change accuracy, and the ones that do not.
Subtitles for video (SRT and VTT)
Which of the two to use, how to load them into an editor or YouTube, and how to fix drift.
Plans and prices, in plain numbers
What Free, Standard and Premium cost, what each one lifts, and how the paid trial works.
Summaries, key points and insights
What each tab above the transcript produces, and when it is worth using.