Enum DiarizationStatus

• Defined in File transcript_result.hpp

Enum Documentation

enum class mosaic::DiarizationStatus

Why a transcript did or didn’t get speaker labels.

Mirrors DIARIZATION_STATUSES in analysis/diarize/pipeline.py; the two must be kept in step. Exists because “diarization: false” alone is unactionable: no Hugging Face token, a gated model refusing to load, and a model that ran and heard nobody all produce a transcript with every speaker empty, which renders identically — a blank Speaker column and an unshaded waveform. The reason used to live only in the run’s stdout, which has scrolled away by the time anyone wonders.

Values:

enumerator Unknown

No status key in the file: written before this field existed. Falls back to has_diarization() for wording rather than guessing a reason.

enumerator Ok

enumerator SkippedByUser

“Transcript only” was ticked.

enumerator NoToken

No Hugging Face token was supplied.

enumerator LoadFailed

The pipeline would not load — usually gated-model terms.

enumerator RunFailed

It loaded but threw while diarizing this file.

enumerator NoTurns

It ran and heard nobody.

enumerator NoOverlap

It heard people, but none of their turns overlapped a transcribed segment. Distinct from NoTurns because the cause and the advice differ: the microphone did capture speech.