Vox Text Overlays
On-screen words are scarce. Each one does exactly one of three jobs or gets cut: (1) drive the eye — ≤3-word anchor + arrow; (2) carry ONE promoted sentence per act — the open loop, staircased; (3) name things — credits, labels, reveals. Never caption the VO in display type; the rail does that.
Highlights animate ON — global rule, no exceptions
The motion of marking is the meaning; a hard-appear reads as pre-printed.
- Yellow text highlight: bar sweeps L→R behind the words.
background-image: linear-gradient(#FFE619,#FFE619); background-repeat: no-repeat;tweenbackgroundSize"0% 100%" → "100% 100%", ~0.4s, stepEase-quantized. - Extend-then-type (the measured source move): the bar grows in steps LEADING the letters by ≤1 step; letter groups toggle in behind it.
- Hand-drawn circles/ellipses: stroke draws on via
stroke-dasharray/dashoffsettween, ~0.4s. - Underlines run L→R. Lower-third role chips with highlight fills count as highlights and sweep too.
- Gate: grep the comp for
.hl,#FFE619, and stroked ellipses — every hit needs an on-animation.
Anchors and arrows
- Anchor = bold small-caps ≤3 words ("THIS FONT", "THAT'S HER", "THIS WORD"), paired with a curved 2-3-step drawn arrow whose TIP touches the subject.
- An arrow without text has no point — always label it.
- Show-before-tell: when the VO is slow to arrive, the anchor may land ~1-1.5s BEFORE its words are spoken; reactions may lead causes slightly, never lag them.
Promoted sentences (staircase text)
One per act, maximum. Staircase layout across 2-3 lines; words step in fast on their own cadence (do NOT wait for slow VO); mixed per-line treatments (letterspaced on a newsprint strip / plain heavy / pencil-underlined / highlight-typed). The whole block settles 1.24→1.0 stepEase power2.out over the beat.
Credit slot
Left-margin, right-aligned: process-blue bold line (#66CFFF) over a small black caps line. Used for photo credits, nicknames, sign callouts, editor credits — same slot every time so the eye knows where names live.
Caption rail
Quiet verbatim rail: 30px 600-weight #222, bottom-center (y≈980), white glow
for dark/blue sections, single-frame HARD cuts between cues (no fades).
Caption the PERFORMED read verbatim. Suppress cues wherever a promoted
sentence already carries those words on screen. If only part of a cue is
promoted, split the cue at a phrase boundary and suppress only the duplicated
part. Display conventional spellings such as 1868 or ATM while preserving
the performed wording and timing. Keep other labels and credits out of the
caption rail's bottom safe zone.