Soma predicts the average viewer's second-by-second brain response with Meta's TRIBE v2, then reads an attention arc and a coarse affect arc off it — no panel, no survey, in minutes.
Primary visual cortex. TRIBE v2 predicts the second-by-second cortical response straight from the video file — benchmarked against real fMRI.
The face- and voice-sensitive cortex — the wedge. A brain-grounded read carries signal a loudness-and-cuts detector can't, and that an LLM role-playing a viewer only guesses at.
A moment-to-moment attention arc: bright where it holds, dimming where attention leaks, with weak-spot callouts pinned to the timeline.
Orbitofrontal and insular cortex read out valence and arousal — pleasant vs unpleasant, calm vs excited — beat by beat.
See the live read-out ↓Drop a video (mp4 / mov / webm, ≤ 150 MB). We run it through the brain and send back the arc — concierge for now, minutes not days.
Overlay two predicted attention arcs on one timeline — the read most performance teams actually want. Test every cut, not just the one you can afford to panel.
Attention and 2D affect ship today. Named emotions ("amusement", "tension", "warmth") come next, as Soma learns from the ads and real audience reactions its design partners generate.