Skip to navigation
Speech to TextFeatures

Emotion detection

Capture per-emotion confidence scores from Pulse STT responses.

View as Markdown
Pre-Recorded

Enabling emotion detection

Include emotion_detection=true in your Pulse STT query parameters.

Sample request

# Download sample audio (or use your own file)
curl -sL -o audio.wav "https://github.com/smallest-inc/cookbook/raw/main/speech-to-text/getting-started/samples/audio.wav"
curl --request POST \
  --url "https://api.smallest.ai/waves/v1/stt/?model=pulse&language=en&emotion_detection=true" \
  --header "Authorization: Bearer $SMALLEST_API_KEY" \
  --header "Content-Type: audio/wav" \
  --data-binary "@audio.wav"

For emotion detection on pre-recorded audio, use Pulse (?model=pulse, emotion_detection=true) - the flat 5-class shape documented below (no neutral). For emotion detection on streaming audio, use Pulse 2.0 (?model=pulse-2, emotion_detection=true by default) - a differently-shaped emotion object per final, with 6 classes (adds neutral), a confidence score, and the full score distribution.

Output format & field of interest

The response adds an emotions object containing floating-point scores (0–1) for happiness, sadness, disgust, fear, and anger. Use these fields to monitor sentiment, trigger QA alerts, or enrich customer analytics.

Sample response

{
  "transcription": "Hello world.",
  "emotions": {
    "happiness": 0.80,
    "sadness": 0.15,
    "disgust": 0.02,
    "fear": 0.03,
    "anger": 0.05
  }
}