Multi-Speaker Podcasts

Overview

Multi-speaker podcasts feature overlapping, informal discussion between several voices. They are useful for diarisation, conversational speech recognition and models that follow multi-party dialogue.

Separate microphone tracks per speaker greatly improve training value.

Key highlights

Reference details

Representative fields a multi-speaker podcasts record may include. Exact fields depend on each owner's systems.

Example record fields

Duration

Episode length

Speakers

Number of voices

Tracks

Mixed or per-speaker

Transcript

Speaker-labelled, if available