Skip to main content
Speech

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

Featurefeatures valueAdds to the response
Summarizationdeepgram-summarizesummary
Topic detectiondeepgram-topicstopics
Sentiment analysisdeepgram-sentimentsentiments
Intent recognitiondeepgram-intentsintents

English only. Audio Intelligence is available on the Starter plan and every higher plan, 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 [email protected] \
  -F features=deepgram-summarize,deepgram-sentiment

Parameters

Send as multipart/form-data.

ParameterTypeRequiredDescription
filefileYesAudio file (WAV, MP3, etc.)
featuresstringNoComma-separated list of the features values above. Default deepgram-summarize
modelstringNoTranscription 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}
  }
}
FieldDescription
request_idID for this request
featuresThe features that ran
transcriptTranscript of the audio
summary / topics / sentiments / intentsPresent only for the features you requested
metadataPer-feature token counts (*_info), the basis of the analysis charge

Pricing

Two parts, charged together on each request:

  • Analysis: billed per token, summed across the features you request: $0.0003125 per 1K input tokens plus $0.000625 per 1K output tokens.
  • Transcription: the audio duration, billed at the per-hour rate of the matching deepgram-* STT model. For example, model=nova-3 bills at the deepgram-nova-3 rate ($0.325 / hour).

See Models.

Errors

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

StatuscodeWhen
400invalid_request_errorfeatures is empty or contains an unknown value. The message lists the valid values
400invalid_requestfile is empty
402insufficient_creditsCredit balance is exhausted
403permission_deniedThe key does not have the stt permission, or a feature is not available on your plan
404model_not_foundmodel is not an accepted model name
413file_too_largefile is larger than 25 MB
429quota_exceededPlan usage limit reached
503provider_error, upstream_timeout, upstream_unavailableThe analysis failed. The response includes a request_id

On this page