OptionaldetectorWhere the speech detector's model and runtime come from (see createSpeechDetector).
Your captions endpoint — your own server, which calls whichever speech service you use and keeps its key (see CaptionsEndpointReply for what it receives and answers).
OptionalfetchFor tests / custom transports. Default: the global fetch.
Optionalinit: RequestInitOptionalinit: RequestInitOptionalheadersExtra request headers (e.g. your app's session token).
OptionallanguagesLanguages your service hears (ISO 639-1), plus 'auto' if it can detect the language; the
Captions panel offers exactly these. Default: the panel's own list and Detect.
OptionalparallelPieces sent at once on long videos (one while the service is busy). Default 3.
OptionalretryFirst wait before trying a busy service again (ms), doubling each time, four tries more;
a Retry-After header from your endpoint wins. Default 3000.
OptionalspeedRoughly how many seconds of speech it captions per second, for "about N min". Default 8.
OptionalvoiceHow to find where people speak in a piece. 'detector' (default where workers run): a small
speech detector on the device (createSpeechDetector) — tells speech from music and noise;
'loudness': by loudness only (no download); or your own (one byte per 20 ms, 1 = speech).
Options of createCaptionsEndpoint.