Expression APIs
Turn human expression into measurements your AI can use.
Measure vocal and facial expression to guide live interactions and prepare recordings for evaluation, training, and analysis.
Voice in
Text in
- 00:04I already tried resetting it twice.
- 00:07Honestly, I just need this working today.
Video in
Scores out
Audio
- expression tags
- 414
- vocal characteristics
- 190
Video
- expression categories
- 48
- visible facial descriptions
- 27
Measured natively in 50+ languages
English, Spanish, Mandarin, Hindi, Arabic, Portuguese, French, German, Japanese, Korean, Indonesian, Bengali, Turkish, Vietnamese, Italian, Polish, Dutch, Thai, Tagalog, Swahili, Urdu, Ukrainian, Malay, Romanian, Greek, Czech, Hebrew, Swedish
Measure expression from voice, face, and text.
Turn vocal, facial, and language expression into structured scores to guide live responses, analyze recordings, and prepare data for evaluation and training. Measure vocal expression, speaking styles, and vocal qualities across 50 languages, alongside facial expression and meaning conveyed through text.
Segment 00:08 – 00:12
- Interest
- 0.52
- Joy
- 0.32
- Mumbling
- 0.13
- Monotone
- 0.07
- Confusion
- 0.02
Frame 00:10
- Concentration
- 0.44
- Interest
- 0.29
- Doubt
- 0.16
Guide live responses. Learn from recorded conversations.
Stream audio for real-time expression scores that inform your agent’s responses, or analyze individual clips for segment-level scores and tags for evaluation and training.
Audio arriving
Scores update inside the turn as the stream continues.
Live scores
- Interest
- 0.35
- Hesitation
- 0.28
- Warmth
- 0.20
- Tension
- 0.08
Go from expression tags to the source audio.
Timestamped results link expression scores to the corresponding speech segment. Review the source alongside its measurements to understand what happened and decide what to investigate next.
Annotated timeline
Select a segmentSelected · 00:02 – 00:06
“I’ve been on hold for twenty minutes.”
- Frustration
- 0.68
- Tiredness
- 0.34
- Interest
- 0.12
Used for
Give your application emotional intelligence.
Robotics
Measure expressions so robots can adjust how they interact with people.
Interviews
Measure vocal expression across candidate conversations, highlighting moments for closer review.
Market Research
Explore how respondents express reactions to products, concepts, and questions, adding context to their answers.
Coaching
Find moments when a sales call changes tone, helping teams review their approach and improve future conversations.
Agents
Use live expression measurements to guide a voice agent’s next response, tone, or pace.
Accessibility
Translate vocal expression into visual indicators and support flexible pacing for different communication needs.
Top-performing models. Grounded in human judgment.
Extensively evaluated against human judgment, our models deliver state-of-the-art vocal and facial expression measurement.
Leading audio performance
Highest overall AUC among 11 systems tested across 32 public datasets.
Leading facial performance
Higher agreement with expert intensity ratings than all compared models on the tested synthetic-face dataset.
Extensive training data
260,000+ speech recordings and 370,000+ facial images from a collection with over one million human ratings.
See what expression adds to your application.
Try an audio example and follow how expression scores change across the sample. Then talk to Hume about your use case, the measurements you need, and how processing fits your workflow.