Speaker diarization is splitting an audio recording into parts by who is speaking, so that a transcript can label each part, for example Speaker 1 and Speaker 2. It answers who spoke when, not what was said. It usually names speakers by number, not by their real names.
Transcribe a recording
Send your voice message, audio or video to @potok_now_bot on Telegram and the text comes back in the chat.
Open @potok_now_bot in Telegram
Files up to 2 GB · export to TXT, MD, DOCX, PDF, SRT and VTT · speaker labels and timestamps · about 30 languages
Key facts
- Diarization answers who spoke when. Transcription answers what was said. A labelled transcript uses both.
- Labels are numbers such as Speaker 1, because the system separates voices but does not know who the people are.
- Similar voices, interruptions and noisy rooms are the usual causes of wrong labels.
How it works in practice
A system listens to the recording and finds the points where the voice changes. It then groups the parts that sound like the same person and gives each group a label. The result is a list of turns such as "Speaker 1 from 00:00 to 00:12, Speaker 2 from 00:12 to 00:25". A transcription tool combines this list with the recognised text, so each passage starts with a label.
What a labelled transcript looks like
[00:00] Speaker 1: Let's start the meeting.
[00:04] Speaker 2: First, we'll review last week's results.
The labels are numbers. They do not carry names, because the system does not know who the people are. You can add names in an editor afterwards.
Why it is not perfect
Diarization tends to go wrong when voices are similar, when people interrupt each other, and when the room is noisy. A short remark from a third person can end up attached to the wrong speaker. For interviews and meetings, read the transcript with the audio and fix the labels. The guide to transcribing an interview shows a routine for this.
Diarization in Potok
Potok can add speaker labels and timestamps to a transcript. You turn them on, and the labels appear as numbers. The text can then be exported as TXT, MD, DOCX, PDF, SRT or VTT. Subtitle files such as the SRT file hold times for each line, while labels help the reader of a document.
Related guides cover Telegram voice messages and MP3 files.
Frequently asked questions
What is the difference between transcription and diarization?
Transcription writes down what was said. Diarization works out who said each part. A transcript with speaker labels uses both.
Does diarization know the names of the speakers?
Not by itself. It separates voices and gives them labels such as Speaker 1. A person adds the real names afterwards.
Why does diarization get things wrong?
Similar voices, people talking at the same time and noisy rooms make it harder to tell speakers apart.
Does Potok label speakers?
Yes. Speaker labels can be switched on. They are numbers, such as Speaker 1 and Speaker 2, not real names.
Sources
Each source was opened by hand on the date shown. Pages change, so open the source if a detail matters to you.
- Speaker diarisation (Wikipedia), Wikipedia · checked
- Speaker recognition (Wikipedia), Wikipedia · checked
Try it on your own recording
Send your voice message, audio or video to @potok_now_bot on Telegram and the text comes back in the chat.
Open @potok_now_bot in Telegram