Speaker diarization and identification in transcription

What is speaker diarization?

Speaker identification in transcription uses diarization to figure out who spoke when. Speech diarization groups passages that sound like the same voice and assigns a consistent neutral label within the recording. Instead of one block of text, you get a conversation organized into speaker turns with timecodes. That makes an interview, meeting, recorded call, or multi-guest podcast easier to follow and correct.

Automatic speaker recognition or speaker diarization?

The terms are often mixed together, but they answer different questions. Diarization separates recurring voices and marks when each one speaks. Speaker recognition or speaker ID can mean matching a voice to a known identity. Hushscript performs the first job: it creates neutral labels from the recording, with no authentication, no biometric identity database, and no guessing at real names.

Online speaker labels: preview first, edit in the app

The public page gives you an online speaker-diarization preview from a file, a supported audio link, or an in-browser recording. It uses automatic language detection and shows up to 5 minutes of speaker-labeled text. The preview has no manual source-language or speaker controls. Those controls, full-recording transcription, and label editing are available after sign-in in the app.

Correct speaker labels without retranscribing

Automatic speaker labels are a starting point. To identify the speaker behind Speaker 1, rename the label once you recognize the person. Then reassign an individual turn to an existing or new label, and split or merge transcript turns when their boundaries need correction. For a solo recording, turn speaker detection off. For recordings that already place speakers on separate audio channels, multichannel handling can use that structure. It still produces transcript labels, not isolated voice tracks.

For a step-by-step walkthrough, see how to fix speaker labels.

Speaker labels in transcription and exports

A useful label tells a reader who owns a passage without pretending to know more than the recording reveals. Keep Speaker 1 and Speaker 2 when anonymity matters, use names when you know them, or choose role labels such as Interviewer and Participant. The transcript keeps labels with their speaker turns and timecodes, and exports represent them in the layout supported by each format. One optional free Insights run creates the complete result together, including speaker topics and name suggestions. Insights stay tied to this transcript version; if you later change the labels, a new generation uses minutes. Check the labels before exporting documents, subtitles, spreadsheets, or editor timelines.

Speaker labels are included with transcription

Speaker detection adds no minutes to a transcription job. Personal accounts buy prepaid minutes only when needed and have no subscription. Hushscript for Teams is a separate monthly or annual service with pooled hours and shared workspaces. In either workflow, the labels organize text; they are not a paid identity-verification service.

Interviews, meetings, and podcasts all read better labeled

Anywhere more than one person talks, speaker labels make the result easier to navigate. In a one-on-one interview, they organize the back-and-forth between interviewer and subject. In a meeting, they help you review who committed to what. In a multi-guest podcast, they help you trace a quote to the labeled turn and timecode before using it in show notes. → Podcasts · Audio to text

Private by design

For local video, Hushscript prepares the audio track in your browser, so the source video stays on your device. Private transcription explains retention, private mode, and what Hushscript removes after processing.

Updated: October 20, 2026