Speaker diarization splits a recording into segments and labels who spoke each one. Speech recognition tells you what was said; diarization tells you who said it.
The system never learns anyone's name. It measures the sound of each voice — pitch, timbre, resonance — and turns short windows of audio into numerical fingerprints. Fingerprints that cluster together belong to the same person. The output is a timeline: one speaker from 0:00 to 0:14, another from 0:14 to 0:31.
Without that timeline a transcript is one undifferentiated block: a question and its answer sit in the same paragraph with nothing to separate them. Attribute each line and the transcript becomes a conversation you can search. You can ask who agreed to something, not just whether it came up.
Accuracy drops when people talk over each other, or when two voices sit in a similar register. A single-word interjection is easy to miss entirely. And if someone rejoins a long meeting from a different device, the system has to decide whether an unfamiliar voice is a new person or a familiar one on a new microphone.
Xtify identifies speakers by their voice, not by which microphone or device picked up the audio. Two people talking into the same mic in a conference room come out as distinct, correctly labelled speakers. Tools that split by device channel cannot do this — both voices arrive on the same channel.
Diarization runs in the same pass that transcribes the meeting and extracts each decision, action item, risk, and follow-up. The labelled transcript goes into a memory index on your Mac. Xtify's speech and language-model providers work under enterprise agreements that exclude your data from model training.
No. Diarization separates voices without knowing whose they are: it can tell you two different people spoke without identifying either. Voice recognition, or speaker identification, matches a voice to a known person. You need diarization first, and most of the time you never need the second step.
No. Because Xtify works from the voices themselves, a single shared microphone in a room still produces separate, labelled speakers. Separate microphones make the problem easier, but they are not a requirement.
Yes. Diarization operates on the sound of a voice, not on the words, so it does not depend on the language being spoken. Xtify transcribes in 60 languages, and on Pro Max and above it follows a speaker who switches language mid-sentence.