Features

Summarization

Turn long recordings into abstracts, bullet lists, and action items.

Summarization condenses an hour of audio into something a teammate reads in a minute. Enable the summarization flag when you create a transcript and the completed transcript comes back with a ready-made summary. When you need a specific shape instead, generate one with a prompt over the completed transcript.

Enable summarization at creation

Set summarization: true on the transcript request to run a summarization pass during processing.

Create a transcript with summarization

curl -X POST https://api.realtimevoicekit.com/v2/transcripts \
  -H "Authorization: Bearer rtvk_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{
    "audio_url": "https://example.com/all-hands.mp3",
    "summarization": true
  }'

When the transcript reaches completed (poll GET /v2/transcripts/{transcript_id} or receive the webhook_url callback), the response includes a summary string. The field is null when summarization was not requested, and also when the pass produced nothing to summarize.

Completed transcript (excerpt)

{
  "id": "tr_abc123",
  "status": "completed",
  "text": "Thanks everyone for joining. Let us lock the launch plan first...",
  "confidence": 0.94,
  "audio_duration_sec": 1860,
  "summary": "The team agreed to ship the beta on Friday, with Dana owning the release checklist and Sam moving the retro to Tuesdays.",
  "chapters": null,
  "entities": null,
  "sentiment_results": null,
  "content_safety": null,
  "created_at": "2026-07-22T14:00:05Z",
  "completed_at": "2026-07-22T14:03:11Z"
}

Generate a summary with a prompt

The built-in summary field gives you one default shape. When your product needs a specific style, ask for it through POST /v2/speech-understanding: the same completed transcript can be summarized many times in different shapes.

bash

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",
    "prompt": "Summarize this recording in five bullet points a teammate can read in under a minute."
  }'

Summary style	Prompt pattern
One-line headline	Summarize this recording in one sentence of at most 20 words.
Executive abstract	Write a three-paragraph summary for an executive who did not attend. Lead with outcomes.
Bullet digest	Summarize this recording in five bullet points a teammate can read in under a minute.
Action items	List every commitment made in this recording as action items with an owner where one is stated.

Structured summaries with JSON output

Set response_format to "json" and describe the object you want in the prompt. The response field then contains a JSON string you can parse directly into your product.

Structured summary

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 a JSON object with keys \"headline\" (one sentence), \"summary\" (three sentences), and \"action_items\" (array of strings)."
  }'

Good to know

The transcript must be completed and contain text; otherwise the endpoint returns 409 with code not_ready.
Prompts can be up to 8,000 characters, so you can include detailed formatting and tone instructions.
Answers are grounded in the transcript. Content that was never said will not appear in the summary.