Proof of Impact

Create an aggregate impact report

Add 2 or more videos of the same event (different angles). We fuse them into one aggregate, audience-level read — energy arcs and indices across the room, never per-camera or moment-by-moment detail. When it’s done you’ll land on a shareable, downloadable report.

Add your footage

Upload a video file or paste a YouTube link. The AI needs at least 30 seconds of footage with a visible face to track 468 facial landmarks and compute behavioural scores.

How your recording is processedWhen you upload a video for analysis, the file may be processed in either of two ways depending on its size, format, and your device's capabilities.(1) On-device path — the video stays in your browser; only derived behavioural statistics (facial landmark summaries, blink rates, expression intensities) are sent to our server. The video itself never leaves your device.(2) Server-processing path — for larger files or unsupported formats, the video is uploaded to a transient processing server, where we extract behavioural statistics plus voice-derived features (pitch, timing, voice quality) and lexical features from the spoken words (e.g. how often you hedge, the variety of vocabulary, sentence structure). The audio waveform and full transcript are processed in memory and discarded immediately afterwards. What we retain in your report is the derived statistical features and a small set of short quoted excerpts (each ≤ 200 characters, capped at 30 per analysis) used as evidence in findings. The original video file is deleted from the processing server within minutes.(3) Cardiovascular analysis (optional, opt-in below) — when enabled, heart rate and heart rate variability are estimated from subtle skin-color variations in the video, a technique called remote photoplethysmography (rPPG). The video is processed transiently on our analysis server; only the derived statistics (heart rate, variability metrics, quality flags) are returned to your device and stored. The video itself is discarded after processing.rPPG is research-grade, not clinical-grade. Do not use these readings to diagnose or rule out cardiovascular conditions. For medical decisions, consult a qualified clinician using validated equipment.Skin tone affects rPPG accuracy: green-channel methods perform best on lighter skin tones (Fitzpatrick I–IV) and show reduced accuracy on darker skin tones (Fitzpatrick V–VI). When reduced accuracy is detected, the report flags it explicitly.Read the full data-flow notice.
CACHING MODELS...
UPLOAD A FILE
Drop video here or click to browse
MP4, MOV, AVI, WebM, MKV · Up to 2 GB · Auto-converts if needed
OR
PASTE A YOUTUBE URL

Upload

Select your video file or paste a YouTube URL

Configure

Choose analysis depth, focus area, and context

Get Results

Full behavioural intelligence report from any video