POST /audio2dialog

POST/third-party/note-taker/audio2dialog

Transcribes an audio file into speaker-labeled segments. The task runs asynchronously — the response returns a task_id immediately and the result is retrieved by polling the status endpoint.

Request

Content-type: multipart/form-data

Field	Type	Required	Description
`audio`	binary	Yes	Audio file to transcribe. Max 300 MB. Supported formats: mp3, mp4, wav, m4a, ogg, flac, aac, opus, webm.
`speakers`	JSON string	No	Array of `{"name", "gender"?}` objects providing known participant names. `gender` is an optional hint with enum `"male"` \| `"female"` \| `null`; omit or pass `null` when unknown. Example: `[{"name":"Alice","gender":"female"},{"name":"Bob","gender":"male"}]`
`contextText`	string	No	Free-form context such as a meeting agenda or domain glossary. Improves term recognition and speaker attribution.
`languageHint`	string	No	ISO 639-1 language code (e.g. `vi`, `en`, `ja`). When omitted, language is detected automatically.

Speaker hints

The speakers field is the most effective way to get correct names in the transcript. Provide it whenever you know who will be speaking. Without hints, segments are labeled speaker_1, speaker_2, etc.

Example

curl https://sandbox.lingint.ai/api/v1/third-party/note-taker/audio2dialog \
  -H "X-API-Key: lk_..." \
  -F "audio=@meeting.mp3" \
  -F 'speakers=[{"name":"Alice","gender":"female"},{"name":"Bob","gender":"male"}]' \
  -F "languageHint=vi"

Response

202 Accepted

{
  "ok": true,
  "data": {
    "task_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890"
  }
}

Poll the status endpoint to retrieve the dialog once the task completes. See Polling task status.