Sign in
assistants

Changed on

Transcribe pre-recorded audio through Vellum's managed Deepgram speech-to-text proxy. Vellum validates the request, checks billing eligibility, calls Deepgram with managed credentials, bills the transcribed duration, and returns the transcript. Pass 'language' to name the spoken language (a BCP-47 code, or 'multi' for code-switching); omitting it decodes as English rather than detecting the language.

post/v1/assistants/{assistant_id}/managed-speech/stt/transcribe/

Request

  • The document declares no server URL.
  • Auth: API key in header Authorization

Path parameters

assistant_idstring uuid required

Request body

audioBase64string required
mimeTypestring required
sourcestring
languagestring

Response

textstring required
providerIdstring required
modelstring required
durationSecondsnumber double required

Changes

  • 8b022ce74761Revision changes
    • ○

      added the new optional request property (media type: application/json)

    • ○

      added the new optional request property (media type: application/x-www-form-urlencoded)

    • ○

      added the new optional request property (media type: multipart/form-data)