Onboarding Plan

Training

# Letterfall Sample — Onboarding Plan A six-scene 16:9 deck recreated from a screen recording. One device carries the whole template: **every headline arrives letter by letter**. - **The letterfall.** Each headline is a `text` layer whose `part: "char"` tween drops the characters in from above — `yPercent: -140`, a 12° tilt and zero opacity, landing on `back.out(1.7)` with a 0.03s stagger. That is the source's one idea: no slide, no wipe, no word stagger, nothing per line. It reads as a scatter settling into type because the tilt unwinds at the same time as the drop. `<br>` is what breaks the lines; the stagger walks the characters in reading order across both, spaces included, so a longer line pushes the tail of the animation later. - **Everything else is still.** Plates and photographs are simply there at the cut; only the two small notes fade, and only after their headline has landed. Nothing is anchored to the scene's end — the recording fades its type out character by character before each cut, and that motion is deliberately not ported, because a filled scene is whatever length its narration turns out to be. The device is worked six ways: a two-line title on a white plate under a photo band (`cover`), a tight two-line greeting straight over a full-bleed still (`welcome`), a three-line argument in a panel beside a half-frame photograph (`why`), a two-line section title in the open corner of a photo grid (`steps`), a two-line payoff on a rounded accent card (`benefits`) and a large sign-off on a full accent ground (`close`). **Type.** One face, Manrope, identified by fingerprint rather than by eye. The recording's cap-normalised ink width for "New Employee" is 9.02 and its x-height ratio 0.754; instanced at wght 400 the candidates read 9.25 / 0.750 (Manrope), 9.38 / 0.707 (Hanken Grotesk), 9.35 / 0.684 (Outfit) — the second number is what decides it, and only Manrope sits on the recording's very tall x-height. Sizes come from measured INK WIDTHS, not cap heights: a cap read off a 262px-tall capture is a pixel wide of the truth and a pixel is 3% of it, which is enough to put a headline a line-break out. That gives 155 / 154 / 126 / 135 / 115 / 172px on the 1080-high canvas, each within 1.5% of the recording once rendered. Line heights are measured baseline to baseline, which is why `welcome` sets at 0.87em and stacks almost touching while the rest sit between 1.25 and 1.29. **The frame.** One 32px margin and one 32px gutter, at every ratio. The source's own grid was not even — stored percent makes the two axes different scales, so its apparently matching `1.9 / 1.9` read as 36px at the sides against 20px top and bottom, with a 7.6px seam between the tile row and the card. Every card grid here closes on the same 32px, which is why the accent cards are shorter than the source's. **Three ratios, re-stacked rather than squeezed.** 16:9 is the base, the ratio the recording was drawn at. Portrait and square are laid out for their own canvas: the photography moves to the top, the words stack beneath, and the accent block closes the frame. Type is recomputed per canvas — `font.size` is resolution-independent, so the same stored number renders 1.78× larger at 9:16 — and every size is set from the longest authored line's measured ink so nothing wraps, which is the one fault the lint cannot see. Two decorations are dropped rather than shrunk, both in 1:1: the hexagon cluster, which has no corner to sit in, and the third grid still in `steps`, which leaves that scene two pictures instead of three. **Eight palettes.** Five lights (Amber authored, Sage, Coral, Plum, Sea Glass, Clay), and two darks (Night, Pine Night) that invert `ground` and carry `ink` light — the mark included, since it is drawn in the `ink` value rather than a hardcoded black of its own. **Where it departs from the source.** The photographs are generic placeholders, not the source's footage, and the deer mark is a plain geometric stand-in. The source is a video deck — its picture slots move — and here they are held stills; the three full-bleed ones are marked `broll` so a filled scene cuts through a sequence instead.