Structured Results

PreviousNext

Recorded Speech can return structured outputs alongside the transcript.

Segment list

Each segment's text, timestamps, and confidence. The core structured output.

Response
{
  "id": 0,
  "start_time": 0.0,
  "end_time": 3.2,
  "text": "The weather is nice today",
  "confidence": 0.95
}

confidence is 0 when the upstream does not provide a value; timing is "estimated" on the production batch path. Treat both as metadata, not as a quality guarantee.

Summary statistics

Audio-level statistics:

Response
{
  "summary": {
    "total_duration": 120.5,
    "total_speech_duration": 95.3,
    "overall_speech_ratio": 0.79,
    "num_segments": 45
  }
}

Metadata

Processing metadata:

Response
{
  "metadata": {
    "language": "zh",
    "chunk_count": 3,
    "audio_duration_seconds": 120.5
  }
}

LLM review results

Review is on by default. Omit review (or set review: true) to run correction; set review: false to opt out.

  1. Per-chunk review: each chunk independently corrected
  2. Global review: final correction after aggregation

The response keeps the corrected lines in result.segments and the uncorrected lines in result.raw_segments; correction details are in result.metadata.correction. Reviewed results are also stored in reviewed_result.json in R2.

Not yet stable

CapabilityStatus
Automatic summary (text)Roadmap
Chapter segmentationRoadmap
Action item extractionRoadmap
Speaker diarizationRoadmap

Stability markers

MarkerMeaning
StableShipped. Safe to depend on in production.
PreviewAvailable, but contracts may still change.
ExperimentalMay change or be removed without notice.
RoadmapPlanned, not yet implemented.