Glossary
Diarization
The task of determining who spoke when in an audio recording, segmenting speech by speaker across turns and overlaps.
Accuracy is measured by diarization error rate (DER), which adds missed speech, false-alarm speech, and speaker-confusion time as a share of total speech time. Overlapping speech is the hard part: when two people talk at once, a system must assign the same moment to more than one speaker, and many pipelines still fail there. Per-speaker audio stems remove the problem at capture, which is why buyers ask whether they exist.
Why it matters
Multi-speaker audio with reliable speaker attribution — or separate tracks — is more useful for training than the same conversation as one mixed file.