A · confidence
Every value has a score
Stored with its method and model. Below threshold it stays out of filters until someone reviews it.
237 variables on every sentence, shot and scene. So agents can find, cut and remix footage without anyone scrubbing a timeline.
Levels run from the whole file down to one sentence. Dimensions say what the variable is about. Tap a cell to open those variables.
Most of the picture lives at the shot and the moment, because that is what search returns and what an editor cuts.
Describe a moment the way you would prompt an AI video model, and any gap becomes a generation brief.
founder podcast, UGC selfie, cinematic film
0232who is on screen, look, product, prominence
0311talks to camera, gesture, energy, handling
048location, interior or exterior, time of day
0510background, clutter, subject position
0612shot size, angle, movement, lens feel
076source, key, direction, contrast
085palette, grade, saturation, finish
0929transcript, key phrase, stands alone, music
107tone, emotion, pace, beat timing
Every client file is split into moments and each moment gets 237 variables: what is on screen, where it fits in an ad, and whether you are allowed to run it. Agents search it like a database and generate only the gaps.
Stored with its method and model. Below threshold it stays out of filters until someone reviews it.
A missing release, unknown music or a risky claim keeps the clip out of results until cleared.
Shared with Creative Intelligence, so what wins across 32M+ ads becomes a search on your own footage.
File data on everything, cheap models on every file, deep AI only on shortlisted moments.
texture shot, application area, skin finish, swatch
5fit view, fabric motion, model size label
6dose shown, ingredient callout, disclaimer required
4appetite cue, taste reaction, plating, temperature
5UI screen, feature shown, workflow step, tap
5room type, style, scale cue, material close-up