Transcription

Language hints, automatic detection and how segments settle.

Transcription is always on. This block controls which languages are expected and when a segment is considered final.

Request

json

{  "transcribe": {    "languages": ["en"],    "strict": true,    "identifyLanguage": false,    "displayMode": "sentence",    "segmentation": "balanced"  }}

Fields

Field	Type	Description
`languages`	`string[]`	Language hints, e.g. `["en"]`. Use `["auto"]` for no restriction.
`strict`	`boolean?`	`false` allows languages outside the hints. Default `true`.
`identifyLanguage`	`boolean?`	Tag each final segment with its dominant language. Default `false`.
`displayMode`	`string?`	`classic` uses endpoint detection. `sentence` (default) groups speech by punctuation, speaker changes and pauses; `segmentation` is ignored.
`segmentation`	`string?`	`short`, `balanced` or `complete`. Default `balanced`.

segmentation trades latency against completeness: short settles sooner, complete waits for fuller sentences and punctuates more accurately.