Transcription with Speaker Labels
Turn on speaker recognition, say how many people speak (two to five) and each line of the transcript gets a speaker label you can rename once and show on every line or only when the speaker changes.
Interviews, podcasts, meetings, calls · 120+ languages
Choose how many people speak
Welcome back. Today we have two guests.
Thanks for having us.
Happy to be here.
Click or drag & drop to upload your audio or videoTap to upload your audio or video
MP3, MP4, M4A, MOV, MKV, WEBM, AVI, WAV, FLAC, AAC, OGG, OPUS, WMA, WMV
No credit card required · Stored in the EU · Never used for AI training
Where Speaker Labels Matter
Questions and answers stay apart, so quotes are attributed to the right person.
Decisions are tied to the person who made them, and you can export one person’s lines only.
Labels renamed to participant codes keep identities out of the exported transcript.
How Speaker Labels Work Here
The settings, the limits and the export options in one table.
- Switching it on
- Recognize Speakers at upload, or later from an existing transcript
- Number of voices
- You choose 2, 3, 4 or 5 before transcribing
- Renaming
- Rename a label once; every line from that voice updates
- Fixing a line
- Move a single line to another speaker in the editor
- In the export
- Names on every line, or only when the speaker changes (TXT, DOCX, PDF)
- One speaker only
- Filter the editor to one person and export just their lines
Getting Clean Speaker Labels
What decides how well voices are separated.
Count
Set the real number of speakers
The model splits the audio into exactly the number you choose. Setting three for a two-person call creates a phantom speaker.
Overlap
Crosstalk is the hard part
When two people talk at once, the line goes to one of them. Recordings where people take turns label best.
Similar voices
Check the first minutes
Two similar voices can be swapped at the start. Fix the first lines and rename; the rest usually follows the same pattern.
How to Get a Transcript with Speaker Labels
From recording to finished text in three steps.
Add the file or paste a link, and pick the language.
Switch on Recognize Speakers and choose how many people speak, from two to five.
Give each label a name, then export with names on every line or only at speaker changes.