Elevator / Pitch / Evaluator
A Coaching Instrument · v0.6

Say it
in sixty
seconds.

Target 60–90s
Whisper · transcribe
Claude · text + vision

Record an elevator pitch — audio or video. We score five rubric dimensions and hand back blunt coaching notes line by line. Video adds three visual delivery dimensions — Presence, Eye Contact, Delivery Confidence — coached alongside the score. No cheerleading.

01
Opening & Credibility
Who you are, why listen.
02
Customer Problem
Specific, felt, named.
03
Value Proposition
Mechanism, not slogan.
04
Call to Action
A door the listener can walk through.
00
Timing
Aim for 60–90 seconds.

+ 3 visual dimensions in video mode · coached, not scored

Scored — audio + timing
Dimension
Exceeds
Meets
Developing
01Opening & Credibility Frame
Immediately establishes who they are and what they offer with specific, memorable language.
Establishes identity and offering clearly within the first 15 seconds.
Vague or delayed opening; listener unclear on who is speaking or what is being offered.
02Customer Problem Identification
Names a specific, recognizable pain point with enough detail to signal genuine customer understanding.
Identifies a customer need or problem, even if somewhat broadly.
Generic or missing; pitch leads with the solution before establishing the problem.
03Value Proposition
Clearly and specifically connects the solution to the stated problem; listener knows exactly what is being offered and why it matters.
Solution is present and relevant to the problem, even if the connection could be sharper.
Solution stated but disconnected from the problem, or leads with features rather than outcomes.
04Call to Action
Ends with a specific, natural next step that advances the conversation.
Ends with some form of invitation or next step.
Pitch ends abruptly or trails off with no clear direction.
00Timing
Within the 45–90 second target window.
Within 15 seconds of the target on either side (30–45s or 90–105s).
Significantly under or over — too brief to be credible or too long to hold attention.
Coached, not scored — video mode only

When you record video, four keyframes are read by Claude alongside the transcript. These three dimensions land as delivery coaching; they don't move the score.

Dimension
Exceeds
Meets
Developing
05Presence
Speaker fills the frame purposefully — head and shoulders well-composed, posture upright and engaged, energy reads visibly across multiple frames.
Speaker is properly framed and stably present, even if energy is neutral. Nothing about the visual presentation distracts from the words.
Off-center framing, slumped or closed posture, low visible energy, or framing that crops the speaker awkwardly (chin cut off, room ceiling dominating).
06Eye Contact
Eyes are aimed at the lens in all or nearly all keyframes — gaze direct, brief natural breaks only.
Eyes are camera-aimed in most keyframes, with at most one frame showing a glance away. Reads as engaged.
Gaze frequently away from the lens — reading off a screen below the camera, looking down at notes, eyes wandering. Reads as 'reciting' or 'unrehearsed.'
07Delivery Confidence
Open posture, hands visible and intentionally used (gesturing in frame), no visible nervous tells across the keyframes — the speaker looks like they own the room.
Settled posture, no distracting motion, hands either visibly at rest or gesturing without fidget. The frame doesn't actively undermine confidence even if it doesn't project it.
Closed or guarded posture, hands hidden or fidgeting, visible nervous tells (touching face, adjusting clothing, swaying out of frame). Body undermines what the words are trying to say.
Press to record.
Grant microphone access when asked. Stop any time — we'll only judge what you give us.
Or try a demo:
Length
60–90 seconds · ±15s tolerated
Format
Audio or video · one take
Privacy
Audio to Whisper, transcript to Claude · shared with the tool author for testing and calibration
Video
4 keyframes to Claude for visual scoring · full recording stays in your browser, never uploaded
Scoring
Exceeds · Meets · Developing
Output
Transcript · scores per dimension · coaching notes
Privacy

Nothing is stored on the server. Audio goes to OpenAI Whisper for transcription; the transcript and (in video mode) four keyframes go to Anthropic Claude for evaluation. Both run zero-retention by default — recordings aren’t kept after the request returns. Your history lives only in your browser’s localStorage. An anonymous copy of the transcript and scores (no name, no account, no IP) is emailed to the tool author for calibration; reach out if you’d rather opt out.

Ver 0.6.0 · Built for practice, not performance.
claude-opus-4-7