Creative
Betacreative_stt_transcribe
Start a transcription job for a public audio file in the connected workspace.
Operation
Write. Requires read-and-write OAuth consent.
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
body.audio_url | string | Yes | Public HTTPS audio URL. Keep it available until the job completes. |
body.language_code | string | No | Language hint. Scribe detects the language when omitted. |
body.segmentation.mode | string | No | Use sentence for timestamped sentence segments. |
body.segmentation.boundary_lead_ms | integer | No | Extra time after a sentence’s final word, 0 to 500 ms. Defaults to 70. |
body.mode | string | No | Use async to receive a job immediately. |
idempotency_key | string | Yes | Reuse the key to retry the same input. |
approval_receipt | string | No | Plan receipt to submit after confirmation. |
The server measures the audio and bills per started minute. Download the complete text, timed words, and requested sentence segments from the JSON file in data.result.artifacts. Large transcripts include a short text preview in the job result.
After confirming the plan, resubmit with its approval_receipt. Pass data.job.id to creative_jobs_wait, then retrieve the output with creative_jobs_result.
Sample prompt
Transcribe this Japanese audio into sentences and give me the transcript JSON URL.