Detected automatically
Auto-detect is on by default. Every change of voice starts a new paragraph labeled Speaker 1, Speaker 2 and so on, each with its own color.
Reviewed by David Dat Nguyen, CXO Last updated
Speaker identification
Who said what.
Labeled for you.
Sonix hears when the voice changes and starts a new labeled paragraph, so an interview, a meeting or a hearing reads like a script instead of a wall of text.
Rename Speaker 1 to a real name once and it changes everywhere. Save a voiceprint and the next recording names that person on its own.
30 free minutes on every new account, no credit card.See pricing
On by default at upload
Rename once, applies everywhere
Voiceprints on Advanced and up
Reads like the conversation
Three things happen to speakers on Sonix, from the first upload to the tenth recording of the same voice.
Named once, everywhere
Type a real name over a default label and every paragraph by that speaker takes it. The same name keeps the same color in every transcript.
Remembered by voice
Speaker Voiceprints learn a voice from about five seconds of clear speech and name that person automatically in new recordings.
Different rooms, the same question: who is talking?
How three kinds of teams work with speaker labels.
Journalists
Quotes need a name, and interviews rarely have one voice.
- A two-person interview comes back as alternating paragraphs, ready to quote
- Rename Speaker 2 to the subject's name once; every quote is attributed
- Save your own voiceprint so your questions are labeled on every interview
Researchers
Focus groups and panels, where attribution is the data.
- Each participant's turns in their own paragraphs, exportable per speaker
- Anonymize with Participant 1, 2, 3 and keep the labels consistent
- Speaker labels travel into DOCX, CSV and NVivo exports for coding
Legal teams
Depositions, interviews and hearings, where who said it matters as much as what was said.
- Examiner, witness and counsel separated into labeled paragraphs
- Rename the labels to the parties once and every paragraph carries the right name
- Save voiceprints for colleagues who appear in every recording
Speakers, under the hood
The settings and limits, as they are in the product today.
- Auto-detect at uploadOn by default
- The upload form's "Auto-detect and label speakers" switch is on by default. Turn it off for a single-voice recording such as a lecture.
- Labels and colorsStable colors
- Speakers arrive as Speaker 1, Speaker 2 and so on, each with a stable color. A custom name gets its own color and keeps it in every file.
- Rename once, everywhereCtrl+Enter
- Overwrite a default label and every paragraph by that speaker takes the new name. Ctrl+Enter cycles speakers on a paragraph; the Speakers breakdown window can merge adjacent speakers or clear all names.
- Speaker Voiceprints25 voices
- On Advanced, Pro and Enterprise plans, save up to 25 voices per account from about five seconds of clear speech. New recordings in that speaker's language are labeled automatically; existing files are not relabeled.
- Multitrack recordings2 to 5 tracks
- Upload two to five tracks and name each one; the track name becomes that speaker's label with no detection needed.
- In exportsDOCX, SRT, CSV
- Speaker names are included in DOCX, TXT, PDF and CSV by default, and can be shown, hidden or set in capitals in SRT and VTT subtitles.
Questions teams ask about
speaker identification
How would a research team use speaker labels in a focus group?
Upload a focus group and each participant's turns come back as their own labeled paragraphs. Rename the labels to Participant 1, 2 and 3, or to real names, once per file, then export to DOCX, CSV or NVivo with the labels attached, so coding by speaker starts immediately.
Does Sonix know who the speakers are?
Not on the first recording. It knows when the voice changes and labels each voice Speaker 1, 2 and so on; you give them names. With Speaker Voiceprints, on Advanced, Pro and Enterprise plans, Sonix remembers a saved voice and names that person automatically in new recordings.
Is there a speaker limit?
There is no fixed cap for a single-track recording; Sonix labels each voice it distinguishes. Clear audio with one person speaking at a time gives the cleanest labels, and crosstalk is what hurts them most. Multitrack uploads take two to five tracks, one speaker per track.
Can a newsroom share saved voices across the team?
Yes. Voiceprints belong to the account, so a reporter's voice saved once is recognized on every colleague's uploads. Up to 25 voices per account, each from about five seconds of clear speech in a recording.
What if two speakers get merged or one is split in two?
Fix it in the editor: Ctrl+Enter cycles the speaker on a paragraph, the Speakers breakdown window merges adjacent speakers, and renaming a label applies to every paragraph that carries it. A quiet room and one voice at a time prevent most of it.
Pages around this one
The feature this page belongs to, and its neighbors.
Everything else Sonix does
with the same recording
Get started
Hear it sorted by speaker
Upload a two-person recording and read it as a conversation. 30 free minutes, no credit card.
4.98
10xfaster than real time
30free minutes to start
