OCI Speech Transcription icon

OCI Speech Transcription

Create and optionally wait for an OCI Speech transcription job, then fetch results from Object Storage

OCI Speech Transcription

AI-generated

Summary

Create an OCI Speech transcription job for an audio object in Object Storage. Return its job ID immediately, or wait for it to finish, locate and download its JSON result, and return the parsed transcript data.

Inputs

  • Compartment ID (required) — OCI compartment in which to create the transcription job.
  • Bucket Namespace (required) — Object Storage namespace containing the audio file and receiving the results.
  • Bucket Name (required) — Bucket containing the audio file; the output prefix is also written here.
  • Audio Object Name (required) — Exact object name or path of the audio file in the bucket.
  • Model Type (required) — Transcription model: ORACLE (default) or WHISPER_MEDIUM. ORACLE jobs are assigned the GENERIC domain.
  • Speaker Diarization (required) — Enable diarization. When enabled, the optional Number of Speakers setting is sent if greater than zero.
  • Output Type (required) — Choose JSON to wait for completion and retrieve the result, or Job ID to return immediately after job creation.
  • Options — Optional settings: Language Code (BCP-47); Max Wait (Minutes), the JSON mode timeout (default 30); Number of Speakers (default 0, letting the service decide); Output Prefix (default transcriptions/); Poll Interval (Seconds), used while waiting (the UI default is 5); Strict Options (default true, fails if reported model or language differs from requested options); and Transcript Format (JSON by default, or SRT). SRT is requested as an additional format, but JSON mode still locates and returns a JSON result.

Output shape

One JSON item per input item, with shape depending on Output Type.

Job ID mode returns {"jobId"}. JSON mode returns {"jobId", "objectName", "result", "appliedModelType", "appliedLanguageCode", "jobReportedModelType", "jobReportedLanguageCode"}; result is the parsed JSON object downloaded from Object Storage. JSON mode polls until the job succeeds, fails on timeout or any non-success terminal state, then searches the configured output prefix and job-related subprefixes for a JSON object. The node requires OCI API credentials.

Links

Discussion