We built a model that scores pitch delivery, not just the script, here's what it caught in a real pitch

Following up on our Inter-1 Streaming work (some of you may have seen our earlier post on the hallucination bug we found). This time it's a product demo rather than a research writeup.

The core idea: transcript-based pitch scoring can't tell the difference between a confident claim and a hedged one, because the words on the page can be identical. "We're growing 40% month over month" reads the same whether you believe it or not.

We built a demo that streams video to Inter-1 in real time and scores delivery signals (confidence, hesitation, energy) alongside a content score, each signal tied to the exact moment it happened.

Tested it on my own pitch. Content scored 87. Delivery caught a hesitation landing right on the traction number, confidence at 50, overall dropped to 80.

Read more here: https://www.interhuman.ai/blog/pitch-practice-demo