Skip to main content
Emotion analysis is Dolva’s second core analysis type. Sending an audio recording to the /v1/analyze/emotion endpoint returns a JSON payload describing the emotional signals present in the audio — including detected emotional states and valence (positive, negative, or neutral tone).

​
What It Measures

Dolva’s emotion models analyze the acoustic features of speech — such as pitch, energy, rhythm, and spectral characteristics — that are known to correlate with emotional expression. The models detect:
  • Emotional states — e.g., calm, tense, engaged, distressed
  • Valence — the positive or negative quality of the emotional tone
  • Signal confidence — how strongly the detected emotion is expressed in the audio
All analysis is acoustic: Dolva processes the audio waveform directly and does not require transcription or language-specific models.

​
Common Use Cases

  • Customer experience — Detect emotional shifts in support calls to flag at-risk conversations
  • Wellbeing monitoring — Track emotional patterns over time for individuals in therapeutic or coaching contexts
  • Content moderation — Identify distress or agitation signals in voice-based platforms
  • Research and analytics — Analyze emotional trends across large audio datasets

​
How to Request Emotion Analysis

Send a multipart/form-data POST request to /v1/analyze/emotion with your audio file:
curl
curl -X POST https://api.dolva.ai/v1/analyze/emotion \
  -H "Authorization: Bearer dv-xxxxxxxx" \
  -F "audio=@conversation.wav"
The response is a JSON object containing the detected emotional signals.

​
Relationship to Cognitive Analysis

Emotion and cognitive analyses are complementary. Running both on the same recording gives you a fuller picture: cognitive signals describe how the speaker is processing information, while emotion signals describe how they are feeling. See Cognitive Analysis to learn more.

View the API Reference

Full request/response details for POST /v1/analyze/emotion