Inter-2
Notice more
Get more insight from conversations, in real time.
AI has learned to understand our words, but words have never been the whole story. The way someone hesitates, the confidence behind a statement, a change in engagement or a moment of uncertainty all give us information that we naturally use when trying to understand another person. Inter-2 brings these social signals into AI in real time, making more of the information contained in a conversation available while it is still happening.
The confidence signal on this clip came from one pass through the Inter-2 upload endpoint — POST https://api.interhuman.ai/v2/upload/analyze.
More than what was said
A transcript tells us what someone said, but it leaves out much of what happens around the words. Inter-2 detects social signals such as hesitation, uncertainty, engagement, confusion, agreement and disagreement as an interaction unfolds, giving AI another layer of information to work with alongside language. This matters because the value of a signal is not simply in detecting it. When social information becomes available while a conversation is still happening, it can become part of what a system understands and how it responds. A research system can recognise when a participant becomes uncertain, a sales platform can see where interest changes, and a coaching product can understand how someone communicates rather than evaluating their words alone.
A new source of insight
The opportunity is not any particular signal or industry, but the information that becomes available when AI can look beyond words. Inter-2 gives products another source of insight into the people they interact with, creating possibilities across customer research, sales intelligence, communication coaching and conversational AI.
Social signal detection — EW-F1
Upper-right is better. Log-scale inference speed.
The benchmark reflects the model's ability to deliver this information at the speed required for real-time applications, reaching approximately 47% EW-F1 at around 3.4× reference inference speed.
The result is not simply more detailed conversation analysis. It is the possibility of building products that can make use of information that was previously difficult for AI to access.
Measured by what it lets you build
The value of Inter-2 is ultimately measured by what its signals allow people to understand, build and improve. By giving AI a new set of insights into people and the way they communicate, Inter-2 can help companies get more from conversations they are already having, from uncovering stronger customer insight to creating more responsive AI experiences.