Chapters and Entities

Three structural flags on POST /v2/transcripts help you navigate long recordings: auto_chapters, entity_detection, and speaker_labels. Enable any combination on the same request.

Set auto_chapters: true to segment the recording into topical chapters during processing. The completed transcript includes a chapters array with one entry per detected chapter, carrying its text and timing values. Chapters give a podcast episode, lecture, or long meeting a natural table of contents.

Set entity_detection: true to detect entities such as people, organizations, and locations while the audio is transcribed. The completed transcript includes an entities array with one entry per detected entity, carrying its text, label, and timing values. Entity awareness is the backbone of CRM enrichment, compliance review, and search.

Both result fields are null unless their flag was requested, and a pass that found nothing also reads as null.

Set speaker_labels: true to separate the conversation into speaker turns. This flag has a direct, visible effect on the transcript response: the utterances array lists each turn with its speaker, text, and start and end times in milliseconds, and the response echoes speaker_labels so you can tell how the transcript was processed.

Speaker separation is part of core transcription; see the Pre-recorded STT docs for the full transcription workflow.

The chapters and entities fields give you the raw structures. When your product needs a combined or custom view, for example one navigable outline that merges both, ask for it in a single call to POST /v2/speech-understanding with response_format: "json".

curl -X POST https://api.realtimevoicekit.com/v2/speech-understanding \
  -H "Authorization: Bearer rtvk_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "transcript_id": "tr_abc123",
    "response_format": "json",
    "prompt": "Return JSON with two keys: \"chapters\" (array of objects with \"title\" and \"summary\") and \"entities\" (array of objects with \"text\" and \"type\", where type is person, organization, location, or other)."
  }'