Skip to Content

Transcribing

Speech-to-text (Whisper) — upload audio and get a transcript via async or sync endpoints.

Endpoints

MethodPathDescription
POST/api/transcribe/asyncUpload audio, return task_id
POST/api/transcribe/syncUpload audio, wait for transcript
GET/api/transcribe/statusPoll status and result

See API Reference for details.

Task IDs

Transcribing tasks use prefix stt:.

Audio input

Two formats are supported:

  1. multipart/form-data with file field audio
  2. Raw binary body (any Content-Type)

Maximum file size: 50 MB.

Optional parameters

Pass via query string or multipart form fields:

ParameterDescription
keep_alive_msKeep-alive hint for the node (uint32)
fanoutNumber of nodes in the auction (default 5)

Status values

Same lifecycle as browsing: queued → … → completed | failed.

Completed responses include success, payload (transcript text), and optional error, execution_time_ms, audio_duration_seconds.