Interhuman

Interhuman develops social intelligence infrastructure for AI applications, delivered as a Signals API. Its core product, Inter-1, is described as an omni-modal model purpose-built for social intelligence: it takes a single video stream and processes video, audio and text together in temporal alignment to detect 12 social signals — agreement, confidence, confusion, disagreement, frustration, hesitation, interest, skepticism, stress and uncertainty among them — along with engagement levels.

Rather than returning only a transcript of what was said, the model reports how something was said by combining cues across modalities, for example distinguishing focus from frustration by pairing a furrowed brow with a vocal pitch shift and tense posture. Output is returned as structured JSON containing the signal type, start and end timestamps, a probability level and a rationale naming the observable verbal and non-verbal cues that triggered the detection. The company states the 12 signals are rooted in behavioral science and that it works with psychologists to validate them.

Business model

Developer-facing API: applications send video and receive detected social signals with rationales and confidence scores, accessed via an API key and supporting documentation.

Traction

The company publishes customer stories including Real Talk Studio, which uses the product to build behavioural evidence beyond words for AI roleplay and communication training, an AI mock-interview application adding non-verbal coaching, and a parent-counselling use case applying socially-aware AI coaching.