HyperFrames captions prompt — embedded animated captions

<direction> An 8-second vertical caption film, 1080x1920 vertical, for burning captions into short-form video. Centre: NOBODY READS A WALL OF TEXT. set large, one word at a time turning orange as a voice would speak it, over a dim base. A thin progress bar fills with the line. At the midpoint the line swaps to SO THE WORDS MOVE INSTEAD. with the final word accented, and a caption-style note explains the rules (one idea per line, highlight the spoken word). Banned: more than ~7 words per line, karaoke bounce easing, glows, gradients, scaling the active word (it shoves its neighbours), highlights that lag or lead the beat. </direction> <structure> Beat grid: line 1 words at 0.5s steps from 0.4s. Progress bar fills with line 1. Swap at ~4.35s. Line 2 words follow on the same 0.5s grid. The how-it-works note lands in the final stretch. Hold to 8s. </structure> <build> One HyperFrames composition, seek-driven GSAP. Words are inline-block spans; highlight = colour only. Give spans real side margins. </build> <gotchas> Two real bugs, one cause each: (1) whitespace text nodes between inline-block spans collapse once transforms apply, eating the word gaps — fixed with margins on the spans; (2) scaling the active word moved every neighbour — colour-only highlight killed the shift completely. </gotchas> Before you finish: step through the timeline at five points and fix anything that overlaps, clips the frame, or drifts off the beat grid.