Skip to main content
People don’t always say what they feel. Someone can call your pricing “fine” while their voice tightens and their face says otherwise. Emotion analysis reads all three signals separately, so you catch the gap between what was said and what was felt.
[VIDEO: Loom walkthrough of emotion analysis — the emotion strip on a clip, a theme’s emotion fingerprint, the emotional intelligence panel. ~1 minute.]

​
Why it matters

Transcripts flatten everything to words. But the difference between “resigned about a problem” and “angry about a problem” changes what you should do about it — and the moments where someone’s words and delivery disagree are usually where the real finding lives.

​
What’s in it

Every participant contribution is scored on three independent channels:
  • Words — what they said, read from the transcript in context
  • Voice — pitch, pace, energy, laughter, sighs, vocal tension
  • Face — micro-expressions, brow, mouth, eyes, head posture
Each channel maps to one of seven emotions: happiness, surprise, neutral, sadness, anger, disgust, fear. Low-confidence reads are filtered out rather than guessed at.

​
When the channels disagree

The most useful signal is divergence — when the channels don’t match. Humanize flags recognizable patterns:
  • Smiling through it — the face shows happiness while the voice or words are negative
  • Forced positivity — the words are positive but the voice or face isn’t buying it
  • Mixed signals — channels disagree without a clean pattern
Divergent moments are marked on clip timelines so you can jump straight to them and watch for yourself.

​
Where it shows up

  • Clips — every clip has an emotion strip: a timeline colored by each utterance’s dominant emotion, synced to playback. Hover for the per-channel breakdown.
  • Themes — each theme carries an emotion fingerprint, a chart of which emotions participants showed while expressing it. Two themes with the same participant count can have very different fingerprints — one resigned, one angry.
  • Summary — the emotional intelligence panel shows the emotion distribution for each moment in your guide, so you can see which topics sparked which reactions.

​
Reading it well

  • Treat divergence as a prompt to watch the clip. The flag tells you where to look; the clip tells you what it means.
  • Compare fingerprints across themes, not in isolation. “More anger than any other theme” is a finding. “Some anger” usually isn’t.
  • Expect plenty of neutral. Most conversation is neutral. The peaks are what matter.