Speaker diarization, explained
A transcript of five people talking is only useful when you know who said what. Diarization is the part of the process that answers "who spoke when", alongside "what was said". In IvreetMeet it is built in: every sentence lands with its speaker, and you name each speaker once.
Two questions, answered together
Transcription asks: what words are in this audio? Diarization asks: how many distinct voices are here, and which stretch of time belongs to which? IvreetMeet answers both and merges them, so speaker labels arrive together with the transcript.
How it works
- 1Find the speech. The stretches where someone is actually talking are located in the recording.
- 2Recognise each voice. Every stretch is described by how the voice sounds, so stretches from the same person group together.
- 3Group them. The number of groups is the number of speakers in the meeting.
- 4Merge with the words. The groups are aligned with the transcript's word timestamps, producing text split by speaker.
From Speaker 1 to real names
Nothing in a recording says "this is Dana". The system knows the third voice is consistent throughout; you supply the name once, and from then on the transcript, the summary, the action items and the search results all use it. It takes about fifteen seconds, and it is the difference between a document people read and a document people skim.
For teams that meet every week, organizations can turn on recognition by voice: a speaker you saved once is suggested by name in later meetings, and you confirm with a click.
Tips for perfect separation
- A microphone per participant when possible, as in Zoom, Meet and Teams recordings.
- One at a time. When people take turns, every sentence lands clearly with its speaker.
- In a meeting room, one good microphone in the middle of the table.
- Participants in the instructions. Listing who was there helps the summary write every name right.
What you can do with the result
Once a transcript is split by speaker, new things become possible: searching for what one person said across every meeting, seeing who spoke and for how long, and turning "he promised to send it" into an action item with a name attached. The free plan covers 7 meetings and 120 minutes a month.
Frequently asked questions
Does it know people’s names?
It detects where the voice changes and labels the groups Speaker 1, Speaker 2. You name them once, and the names appear in the transcript, the summary and every export.
How many speakers can it handle?
As many as the meeting has: a two-person call, a team meeting or a panel. Each voice gets its own speaker.
Do I have to tell it how many people are in the room?
No. You are never asked how many people are speaking; the system works it out from the audio.
Will it recognise the same person next time?
Yes, in organizations that turn on recognition by voice: a speaker you saved once is suggested by name in later meetings, and you confirm with a click.
How do I get the cleanest separation?
A microphone per participant when possible — which is what Zoom, Meet and Teams recordings give you — and one person speaking at a time.
Try it on your own recording
IvreetMeet transcribes and summarizes Hebrew meetings. The free plan covers 7 meetings and 120 minutes per organization each month.
Start free