This is a per-letter reveal animation I built entirely with CSS - no JavaScript timing loop, no animation library, just a single custom property doing the heavy lifting.
Each letter gets its own --index, set inline via style={{ '--index': i }} on the span. That value feeds animation-delay: calc(0.03s * var(--index)), so instead of every letter starting at once, each one waits a little longer than the last - letter 0 starts immediately, letter 9 starts 270ms later. That’s what gives you the cascade instead of a flat pop-in.
One detail I almost missed: animation-fill-mode: backwards. Without it, every letter flashes into its final position for a frame before its delay kicks in, which looks broken. With backwards, each letter just holds still - translated out of view - until its turn comes up. The easing, a cubic-bezier(.075, .82, .165, 1) curve, slows down hard right at the end so it settles instead of overshooting and bouncing.
LetterCascade.tsx
import styles from './styles.module.css'const TEXT = 'Animations'export function LetterCascade() {return (<h1 aria-label={TEXT} className={styles.h1}>{[...TEXT].map((letter, index) => (<spanaria-hidden={'true'}key={`${index}-${letter}`}className={styles.letter}style={{ '--index': index }}>{letter}</span>))}</h1>)}
styles.module.css
.h1 {font-size: clamp(32px, 5vw, 48px);font-weight: 600;letter-spacing: -0.05em;overflow: hidden;color: #0a0a0a;}.letter {display: inline-block;animation: yTranslate 1.3s cubic-bezier(.075, .82, .165, 1);animation-delay: calc(0.03s * var(--index));animation-fill-mode: backwards;}@keyframes yTranslate {from {transform: translateY(100%)}to {transform: translateY(0);}}@media (prefers-color-scheme: dark) {.h1 {color: #fafafa;}}@media (prefers-reduced-motion: reduce) {.letter {animation: none;}}
It also respects prefers-reduced-motion - if you’ve told your OS you’d rather skip animations, you just get the final text straight away, no cascade, no fuss.