vertical.md
23 lines · skills/hyperframes/references/types/vertical.md at a595755
Vertical and social clips (9:16)
Made at 1080×1920 when the request or the platform says vertical; a landscape film is not squeezed into it but recomposed.
Raw lines 1 to 4The platform's own UI
The app draws over the frame. On 1080×1920: nothing in the top 168 px (status and search bar); every word and the product inside the band from 168 to 1330 px; below 1330 px (caption, handle, buttons) only texture or decoration; no text under the right-hand action rail. Ads and some apps reserve more at the bottom; check the platform the clip is for. Burned-in captions sit at the bottom of the band (../captions.md).
The first second
- Frame 0 is the thumbnail and the hook: a face, the result, or one bold line already on screen. Never black, never a logo.
- The viewer decides in about a second whether to stay: the change or the question arrives in the first beat.
- Many viewers watch muted: speech is captioned, and the picture carries the hook without sound.
Layout
- One subject, centred, stacked vertically; landscape layouts (side-by-side panels, wide charts) are rebuilt as a vertical sequence.
- Type is larger than in landscape (titles about ×1.35,
../overlays.md); nothing meant to be read under 28 px, and a fixed word limit per line (about 8 Chinese characters or 5 English words for a claim): cut words, never shrink. - Talking heads: punch-ins on emphasis, B-roll cut on the sentence, the face never under a card (
../overlays.md).
Pacing
Faster than landscape: shots of 1–2 s, a new beat every 2–4 s, the payoff before the swipe. A loop that cuts from the last frame back to the first is a choice, not a default.
Raw lines 21 to 23