Audio Intelligence

Summarize an audio file and detect its topics, sentiment and intents in one call.

Overview

Upload an English audio file and get back its transcript plus any combination of four analyses: a summary, topics, sentiment and intents.

Endpoint: POST /v1/audio/intelligence

Feature	`features` value	Adds to the response
Summarization	`deepgram-summarize`	`summary`
Topic detection	`deepgram-topics`	`topics`
Sentiment analysis	`deepgram-sentiment`	`sentiments`
Intent recognition	`deepgram-intents`	`intents`

English only. Audio Intelligence is available on the Starter, Pro and Enterprise plans, and the key needs the stt permission. To run the same analyses on text instead of audio, use POST /v1/text/intelligence — see Models.

Basic Usage

curl -X POST https://api.callmissed.com/v1/audio/intelligence \
  -H "Authorization: Bearer cm_your_key" \
  -F file=@call.wav \
  -F features=deepgram-summarize,deepgram-sentiment

Parameters

Send as multipart/form-data.

Parameter	Type	Required	Description
`file`	file	Yes	Audio file (WAV, MP3, etc.)
`features`	string	No	Comma-separated list of the `features` values above. Default `deepgram-summarize`
`model`	string	No	Transcription model used for the analysis. Default `nova-3`. Also accepts other Deepgram model names such as `nova-2`, `nova-2-phonecall`, `nova-2-meeting` or `enhanced` (the `deepgram-*` STT IDs without the `deepgram-` prefix)

Response

{
  "request_id": "dgai-3f9c1a2b4d5e6f70",
  "features": ["deepgram-summarize", "deepgram-sentiment"],
  "transcript": "Hi, I'm calling about my order...",
  "summary": {
    "result": "success",
    "short": "The customer called to ask about a delayed order."
  },
  "sentiments": {
    "segments": [
      {"text": "Hi, I'm calling about my order...", "start_word": 0, "end_word": 7, "sentiment": "neutral", "sentiment_score": 0.1}
    ],
    "average": {"sentiment": "neutral", "sentiment_score": 0.05}
  },
  "metadata": {
    "summary_info": {"input_tokens": 120, "output_tokens": 28},
    "sentiment_info": {"input_tokens": 120, "output_tokens": 0}
  }
}

Field	Description
`request_id`	ID for this request
`features`	The features that ran
`transcript`	Transcript of the audio
`summary` / `topics` / `sentiments` / `intents`	Present only for the features you requested
`metadata`	Per-feature token counts (`*_info`), the basis of the charge

Pricing

Billed per token, summed across the features you request: $0.0003125 per 1K input tokens plus $0.000625 per 1K output tokens. See Models.

Errors

Errors use the OpenAI envelope: {"error": {"message", "type", "code"}}.

Status	`code`	When
`400`	`invalid_request_error`	`features` is empty or contains an unknown value. The message lists the valid values
`402`	`insufficient_credits`	Credit balance is exhausted
`403`	`permission_denied`	The key does not have the `stt` permission, or a feature is not available on your plan
`404`	`model_not_found`	`model` is not an accepted model name
`429`	`quota_exceeded`	Plan usage limit reached
`502`	`provider_error`, `upstream_timeout`, `upstream_unavailable`	The analysis failed. The response includes a `request_id`