Wrench.ai scores video creative by slicing it into individual frames — typically one or two frames per second — describing each frame, and then scoring each one against your persona meta-measures. The individual frame scores are then aggregated into an overall video score, with heavier weighting on the moments that matter most: the opening hook and the closing brand moment.
One 30-second ad is the equivalent of roughly 30 images being scored. The cost and complexity are directly correlated to the length and sophistication of the video.
The video gets broken into static frames at a rate of one to two frames per second. For a 15-second spot at 1 frame/second, that's 15 individual images. Each frame is described and then scored independently — the same way any static creative would be scored — across brand fit, audience fit, creative novelty, and contextual appropriateness.
This approach captures the full narrative arc of the video. Every beat, transition, and visual moment gets evaluated against the target persona.
For a 15-second spot, 1 frame per second (15 frames) is the right cadence. Every second counts in a 15-second spot — there's no filler to skip. For longer content (30s, 60s), you could go to 1 frame every 2 seconds to keep scope reasonable while still capturing key transitions. The baseline is one to two frames per second depending on complexity.
The individual frame scores are combined into an overall video score using weighted aggregation. The opening frames (first 3 seconds — the hook) and closing frames (last 2 seconds — the CTA/brand moment) receive heavier weighting, because these are the moments with the highest impact on audience retention and action.
The confidence of the aggregated score is directly tied to the richness of the persona meta-measures. The more detailed the persona ingredients, the more precise each frame score — and the more meaningful the aggregation.
Most clients won't have 200 videos to work with — they'll have a few. For that reason, it often makes more sense to maintain a shared workspace (such as the Hispanic workspace) where client videos are added and compared against an accumulated bank of 100+ previously scored videos, rather than trying to build a standalone model for each client.
You're managing their one or two videos against your bank of scored content that's been accumulated over time. This gives every new video a much richer comparison set from day one.
Video scoring is directly correlated to the length of the ad. Each frame is the equivalent of scoring one static image. A 15-second spot at 1 frame/second = 15 image-equivalents. A 30-second spot = 30. Clients can estimate scope that way.
This is a newer product capability and pricing may vary by engagement. The frames are sliced, described individually, and reassembled — so the sophistication of the video (rapid cuts vs. slow pans, multiple scenes vs. single setup) can affect the descriptive complexity.