When a recording has more than one person in it, Fluen can work out who is speaking and mark it in the subtitles. You decide how that looks: named labels, a simple dash on every change of speaker, or nothing at all. This guide covers the three modes, where to switch them on, and how to rename or reassign a speaker in the editor.
How it works
Alongside the transcription, an industry-leading speaker recognition model listens for distinct voices and assigns every cue to the right one. It works on any language Fluen transcribes. Running it costs no credits, and it usually takes seconds (a little longer on files over 30 minutes).
The three modes
| Mode | What the subtitle looks like | Good for |
|---|---|---|
| Off | When we first started looking at the data, the patterns were surprising. | Single-speaker content, voice-overs, lectures |
| Dash | - When we first started looking at the data, - And what did you find? | Broadcast-style subtitles where a dash marks each change of speaker |
| Labels | [Lena Halberg] When we first started looking at the data, | Interviews, podcasts, panels, meeting transcripts |
With Labels, the transcript export (Word, PDF or text) is also grouped by speaker into readable paragraphs.
Where to turn it on
As a workspace default
Open Workspace Settings, then Caption Defaults, and set Speaker Detection to Off, Dash or Labels. Every new upload in the workspace starts with that choice. Details in Set your default subtitle specifications.
For one upload
On the upload screen, expand Subtitle settings and pick Off, Dash or Labels under Speaker detection. This overrides the workspace default for that file only.
On a track that already exists
In the editor, click Speakers in the toolbar above the subtitle list and choose a mode. If speaker detection has not run on this file yet, a dialog called Enable Speaker Detection explains what will happen; click Detect Speakers and wait a few seconds. Tick "Enable by default for new uploads" in that dialog if you want the same setting on future files.
Renaming and reassigning speakers
In Labels mode, every cue in the editor shows its speaker's name on the left, in a colour of its own. The colour is an editor aid only; exports use the name in square brackets.
- Rename a speaker: click the name, type the real one and press Enter. The new name applies to every cue of that speaker, in one go.
- Reassign one cue: click the small arrow next to the name and pick another speaker from the list. The last entry, + New Speaker, creates a speaker who was not detected and assigns the cue to them.
- Cues the model could not attribute are labelled Unknown; assign them the same way.
In your exports
- Subtitle files (SRT, VTT and the others) carry the dash or the
[Speaker Name]label as plain text at the start of the cue. - Burned-in video: the Export Video with Subtitles dialog has its own Speakers control, so you can render a labelled version and a clean version from the same track.
- Transcripts are grouped by speaker when labels are on.