The sound, as 16 kHz mono PCM in order, in pieces of at most 30 s (so memory stays flat on long videos). Pieces follow each other with no gap.
Length of the whole sound in seconds (for progress).
Spoken language as an ISO 639-1 code ('en', …), 'hi-Latn' for Hinglish (Hindi written in
Latin letters), or 'auto' to detect it (one of languages).
OptionalonCancels the transcription; the promise then rejects with an AbortError.
OptionalvocabularyNames and words that may be said (people, brands, places), so they're spelled right — for
transcribers with takesVocabulary. Optional.
What the editor hands a transcriber.