A founder on our team recorded a pitch. Read as a transcript, it was good: a clean hook, a real traction number, a crisp ask — the content scored in the high 80s.
Then we looked at the same recording the way an investor would. Clarity came back an 87. But energy sat at 66 and confidence at 50, dragged down by a beat of hesitation right before the traction number and an ask that lost steam at the close. Strong words, shakier room. The overall came out an 80 — and that gap between the two halves is the whole point of this demo.
TL;DR: We built a pitch-practice demo on Inter-1 Streaming that scores how a pitch is delivered, not just what it says. The script is the easy half to grade. The half that decides the room — confidence, hesitation, energy, presence — is the one nobody was measuring.
What the demo does
The Pitch Practice is a small reference app built on Inter-1. You record, it watches, you get scored on delivery and content together. Two ways in:
- Practice your pitch — record a full 1–3 minute pitch to your webcam and get feedback on confidence, clarity, energy, and the words themselves.
- The 1-Minute Challenge — exactly sixty seconds, auto-stop, with an optional global leaderboard and a percentile so you can see where you land against other founders.
While you record, each few seconds of video streams straight to Inter-1, and the behavioural read comes back in step — so by the time you stop talking, the analysis is already done.
The half nobody scores
Pitch coaching is almost entirely about the script. Tighten the hook. Lead with the problem. Put a number on the traction. Land the ask. All good advice, and all of it about words — the half you can write down and edit in a doc.
But investors don't decide on the doc. They decide on the founder in front of them. And the read happens on a channel the transcript never captures: the half-second of hesitation before a number you're not sure of, the energy that drains out of the ask, the confidence that's there for the problem and gone for the solution. The same sentence — "we're growing 40% month over month" — can sound like a fact or like a hope, and the difference is delivery.
So we put both on one timeline. The words run along the bottom; the behavioural signals — confidence, hesitation, stress, uncertainty, interest — sit on top, each pinned to the moment it happened, each with a reason attached.
Take that pitch from the top of this post. The content scored in the high 80s — the message was clear, the ask was there, the structure was clean. The delivery score told the other half of the story: a hesitation signal landed right on the traction figure, with the rationale "a noticeable pause and a rising, uncertain tone right before the revenue numbers — the traction is strong, but the delivery hedges it," and energy dropped through the close. Read together, the feedback isn't "your pitch is a 9." It's "your words are a 9; now go say the traction number like you believe it."
How the score is built
The headline number is an even split: half delivery, half content. We weight them equally on purpose — a polished script delivered flatly and a shaky script delivered with conviction are both incomplete, and a founder should see which half is letting them down.
Delivery comes from Inter-1 — its conversation-quality read plus the social signals over the recording:
- Authority — decisiveness and credibility (30%)
- Clarity — how easy you are to follow (25%)
- Energy — vitality and presence (20%)
- Confidence — net of confidence signals against hesitation and uncertainty (15%)
- Low hesitation — a penalty for hesitation, stress, and uncertainty piling up (10%)
Content comes from the transcript, scored against a six-part rubric — message clarity, problem framing, solution and value, evidence and specificity, narrative, and the ask — blended with two mechanical speech metrics: pace (around 150 words per minute reads best) and filler density (every "um," "like," and "you know" counts).
None of this replaces a human coach. It's a mirror that shows you the half of the performance you can't see while you're giving it — and, crucially, shows it the way Inter-1 reads a room, not the way a rubric reads a paragraph.
It's a reference implementation on Inter-1 Streaming
Here's the part for the people who build things. Everything above — the live signals, the delivery score, the timeline that updates as you speak — is Inter-1 Streaming doing the work. The app opens a WebSocket, streams the recording up a few seconds at a time, and Inter-1 streams typed events back as signals start and end. No buffering, no waiting for the upload to finish before analysis begins. The behavioural read arrives in step with the speech.
The pitch app is one thing you can build on that layer. The same stream powers a research-interview tool and the live browser demo. If you're building anything where how someone says it matters as much as what they say — coaching, interviewing, sales enablement, support — Inter-1 is the behavioural layer you'd otherwise have to invent. This demo is a working blueprint for plugging it in.
Scoring the words was the easy half. The half that decides the room is the one nobody was measuring.
A transcript scorer was never going to tell a founder why a clean pitch fell flat, because the answer isn't in the transcript. It's in the beat before the number, the energy in the ask, the confidence that's there for the problem and gone for the solution. That half has always been the half that gets you the check — and until you could read behaviour as fast as you read words, it was also the half nobody could put a score on. That's the half this demo is for.
Record a pitch and see your own delivery scored — it takes about a minute. And if you want to build the behavioural layer into something of your own, start with the Streaming API.
