The engagement read, before the views.

You find out whether a clip works after it has gone out. Views, retention curves and ad metrics arrive days later, cost the spend that produced them, and describe an audience average — they say that attention dropped, not which second, not whether it was the picture or the sound, and not what to do about it.

Testing the honest way needs reach most people do not have: an A/B test on a real audience, or a panel with sensors on their heads. So the decision between two cuts gets made the old way — whoever argues best in the edit room wins. And the “AI video score” tools that promise a shortcut hand back a number with nothing underneath it.

NeuroVidz reads a clip the way a viewer's brain would. Every second of picture and sound is distilled into 39 perceptual signals — motion and cuts, faces and expression, colour and light, loudness, voice, music, novelty. Those signals are mapped onto how seven cortical networks respond: the ones that carry attention, salience, emotion and narrative. The mapping is grounded in published functional-imaging research, and the method is open to read.

What comes back is not one number. It is a per-second timeline of the response, a craft score from 0 to 100 with the four factors behind it, an emotion read over time with its confidence at each moment, and the specific seconds — and edits — to act on.

Two design rules hold everywhere. Sound is read as seriously as the picture, because so much of engagement rides on it. And when a read is not clear, the system says “no clear read” instead of inventing a figure — every result is reproducible from the same artifacts, and a number you cannot trust is worse than no number.

Drop a video or audio file — mp4, mov, webm, mp3 or wav, up to 400 MB. Each read covers up to three minutes; for a longer file you pick the section to read, right in the uploader, and only that section is analysed and charged. About thirty seconds later you have the score, the attention curve, the emotion timeline and the recommended edits. Run the next cut the same way and compare.

It is prepaid, not a subscription, and nothing renews by itself: 10 credits read up to a minute of video, or any audio clip up to three minutes, and every new account starts with a few to see a real result first. Your clips and your data can be exported or deleted from your account at any time.

• Performance marketers and agencies. Pre-flight an ad before the media spend. Run every cut of a campaign, keep the hook that holds attention, and put the read in the client review instead of an opinion.
• Creators and editors. Settle which of two edits is stronger, find the second the middle sags, and see whether the opening three seconds earn the rest.
• Podcasters and audio producers. Sound is read on its own terms — voice, music, the beat drop, the held silence — so an audio-only episode gets the same second-by-second attention curve.
• Product and brand teams. Test a launch film, an explainer or a demo before it is published, and compare the version you love with the version that lands.
• Researchers and educators. A reproducible per-second model of how seven cortical networks respond to a stimulus, from the same artifacts every time.

NeuroVidz is a product of Sandmatter LLP, built and run by two data scientists who spent years shipping video-understanding systems at platform scale and wanted the instrument they never had in the edit room. Questions go to real people — talk to us.