Motion Language

A good promo feels like one continuous film — one camera, one motion feel, smooth and timed to the voiceover — not a pile of slides that animate once and freeze. The doctrine in Part 2 is load-bearing: when in doubt, do what it says.

Part 1 — the move vocabulary

Reach into this palette when naming a scene’s motion. Pick the move that matches the beat, name it in the shot sequence, and cite the rule id after →. The blueprints (../hyperframes-animation/blueprints/) name these same moves in their rule mapping; you’re drawing from one shared palette. Compose 2–4 across a shot’s scenes (entrance → sequential reveal → settle), not all at once.

Kinetic type

• hard-cut / flash word-swap — a word or line replaces the previous one on an instant cut (no fade/roll); the swap itself is the beat. → discrete-text-sequence
• in-place token cycle — a fixed line holds and only its variable slot changes, token → token → token. → discrete-text-sequence
• per-word staggered reveal — a phrase assembles word-by-word (or chunk-by-chunk), each landing on its own beat. → dynamic-content-sequencing
• kinetic beat-slam — short phrases slam in on a shared percussive beat array, each with a distinct entrance, resolving on a locked finale; the recipe for “punchy / rhythmic” taglines. → kinetic-beat-slam

Typewriter

• type-on with caret — text types in character-by-character behind a blinking caret. → discrete-text-sequence (+ context-sensitive-cursor for the caret blink / color)
• backspace-and-retype — the line types, deletes the last word(s), and retypes a new one (typo-correction, reframe). → discrete-text-sequence (+ context-sensitive-cursor)

Count-up / data

• value-scaled counter — a number counts up and its font size grows with the value, so the climb itself escalates. → counting-dynamic-scale
• bars / progress / star wipe — a number paired with a graphic that fills: bar-height stagger, a progress bar / ring filling, a fractional star-rating wipe. → stat-bars-and-fills

Reveal / decode

• 3D char flip-decode — characters flip in 3D and resolve from scrambled glyphs to the real text (decryption feel). → hacker-flip-3d
• SVG self-draw — an outline / icon / ring draws itself stroke-by-stroke. → svg-path-draw

Camera

• push / focus / drift — a sequential camera move on the frame root (pull-back → focus → push) plus continuous micro-drift; the cinematic baseline. → multi-phase-camera
• zoom-to-target — zoom into a non-centered element (scale + counter-translate to keep it framed). → coordinate-target-zoom
• pan / focus-lock — a virtual camera transforming one .world wrapper to pan / zoom / lock onto a region. → viewport-change
• camera-cursor-tracking — the viewport locks to a moving focal point (a typing cursor), static framing then focal-locked tracking. → camera-cursor-tracking

Layout motion

• cluster→outward expansion — elements start clustered at center and expand outward to their final positions in lockstep. → center-outward-expansion
• orbit — elements flip in from 3D space and settle into a continuous elliptical orbit (entry flips in-place at the orbital position). → orbit-3d-entry
• split-tilt cards — two cards side-by-side with opposing rotationY tilts, entering from their respective sides (comparison / before-after). → split-tilt-cards
• logo/avatar ring + connectors — avatars or logos on an elliptical ring with SVG connection lines to a center point, staggered entry. → avatar-cloud-network

Surface / UI

• 3D page-scroll reveal — a full webpage as a tilted 3D card whose internal content scrolls to reveal specific sections. → 3d-page-scroll
• cursor click + ripple — a cursor moves to a target, depresses with it on click, and emits an expanding ripple. → cursor-click-ripple
• button press — a tactile press: compression then spring recovery, optional release burst / glow. → press-release-spring (or physics-press-reaction for a click that compresses cursor + target together)
• keyword glow — keywords light up with glow + scale + color on an attack-decay-rest envelope, synced to a word rail. → asr-keyword-glow

Morph / handoff

• scale-swap — two elements at the same screen center hand off: the outgoing cluster shrinks + fades as the incoming one arrives. → scale-swap-transition
• card morph-anchor — a container morphs apparent size + corner radius + surface between two shots, then fades to reveal the real target beneath (HyperFrames uses uniform scale, not width/height). → card-morph-anchor

Seam cuts (worker-built, inside a frame)

The velocity-matched cuts a worker authors between a frame’s own Scenes. Name the seam in the shot sequence; the recipe is in the catalog, not a single ../hyperframes-animation/rules/ id.

• zoom-through / inverse zoom-through — a within-scene swap on the Z-axis; forward reads “progressing through”, inverse reads “arriving at” (payoff). → cut-catalog.md
• cut-the-curve — a scene-to-scene cut where both sides move the same direction at matched velocity. → cut-catalog.md
• waterfall cut — cut-the-curve at word granularity, a wave across a text-to-text seam. → cut-catalog.md

Emphasis / marker

• highlight / circle / burst / scribble — a marker-drawn emphasis on a word or element: yellow highlight sweep, hand-drawn circle, radiating burst, scribble, or rough sketch-outline. → css-marker-patterns

Aliveness during a hold (use sparingly — see Part 2)

• subtle jitter — the sanctioned way to keep a settled frame alive: a small, low-amplitude positional/scale jitter on the held element. The motion-graphics trick that reads “alive” without reading “weak.” → sine-wave-loop (low-amplitude register)
• live SVG internals — internal SVG parts move so an icon feels alive (rotating hands, oscillating blades, pulsing dots, dash-flow); fine because it’s the subject doing something, not a card breathing. → svg-icon-enrichment
• finite bounded ambient — a single bounded breathe/drift on ONE held hero, only when genuinely needed; de-emphasized — prefer sequential reveal or jitter first. → sine-wave-loop

The added moves — now backed by local rules

Five moves the golden corpus needs were added to this skill’s ../hyperframes-animation/rules/, rounding out the vocabulary above:

• depth-of-field / selective-blur — blur the off-focus subset to spotlight the focal element → depth-of-field-blur
• motion-blur streak — directional velocity blur on a fast fly-in / camera push-through → motion-blur-streak
• 3D depth scatter-assemble — glyphs/elements scatter into a tumbling 3D cloud, then reassemble → depth-scatter-assemble
• spring-pop entrance — the canonical entrance pop; default to a smooth long-tail settle, overshoot only when explicitly playful → spring-pop-entrance
• ambient glow / bloom — un-triggered soft glow blooming behind a static hero → ambient-glow-bloom

Part 2 — the motion doctrine (load-bearing)

These four rules are the difference between a clip that reads as a serious launch video and one that reads as an agent-made PowerPoint. Follow them as written.

1. Smooth beats bouncy — power3 is the default

Elements should use long-tail decel curves that let them settle smoothly. power3 is enough in most cases. No bouncy, no overshoot, no back.out / bounce.out / elastic.out as a default.

Bouncy is the #1 instant turn-off in user-made Remotion / HyperFrames videos, and the agent almost never gets it right — it thinks bouncy adds emphasis, but it buys that emphasis at the cost of cleanliness. The serious launch-video shops feel the same. Smooth always wins. Overshoot is demoted to a rare, explicitly-playful exception (a consumer/fun logo slam, a deliberate bell-hit) — never the house style. Name the intent as a long-tail settle; the worker maps power3 (or expo.out on a fast arrival). See ../hyperframes-animation/rules/spring-pop-entrance.md — it now leads with the smooth settle. (The exact form of that settle is a critically-damped spring; the worker has a baked, seek-safe springEase — ζ=1 — in ../hyperframes-animation/adapters/gsap-easing-and-stagger.md → Spring Eases for when the settle is the hero. Real physics, same doctrine — not a license for bounce.)

2. Sequential reveal in the back ~50%, timed to the voiceover

This is the anti-PowerPoint mechanism — sharper than “put development in the middle.”

• Don’t dump everything on screen in the first ~25% of the scene. Rushing all content in up front is exactly what forces the slideshow feel.
• Reveal each piece — a line, a card, even an h1 — when the voiceover mentions it, sequencing reveals across the later ~50% of the scene. Same amount of agent work, but the cut becomes coherent and gains rhythm.
• Less is more. Fewer things on screen, each arriving on its VO beat, beats a full canvas that animated once and froze.

Practically: a frame’s shot sequence front-loads almost nothing — the entrance carries only what the VO is saying at t=0, and the rest of the elements wait in the timeline for their spoken cue. A reveal maps onto a development-class move from Part 1 (per-word staggered reveal, cluster→outward expansion, a count-up, an asr-keyword-glow synced to the word rail).

3. No lazy breathing, no bad pan/push — “no motion over bad motion”

The agent’s two reflexive ways to fake “aliveness” both read cheap:

• No lazy breathing. Scaling cards/text up and down in a circular loop to look “alive” is the cheap tell. Don’t reach for it.
• No bad slow pan / push in the back half. A slow pan or push on elements in the later ~50% of a scene disrupts the viewer’s sightline and causes eye discomfort — it actively makes the frame worse, not better.

The fix for both is the same: stagger element reveals in time with the script (rule 2). And the governing principle: “I’d rather have NO motion than BAD motion.” A held, still frame is better than a frame kept “alive” by breathing or a drifting camera. The only sanctioned aliveness during a hold is subtle jitter — a small low-amplitude jitter that keeps a frame from feeling dead without looking weak (it’s in current video work now). Everything else holds.

4. Internal seams are velocity-matched cuts

When a frame has an internal seam — a within-scene swap, a Scene-to-Scene cut, a text-to-text line change — make it a velocity-matched cut, not a hard slideshow cut: cut at peak velocity, match direction and speed on both sides. The catalog (the four techniques, the blur logic, and which to use when) is cut-catalog.md; the moves are listed under Seam cuts in Part 1.

One-line summary

Smooth long-tail (power3) over bouncy; reveal sequentially in the back ~50% timed to the VO (not dumped in the first 25%); no lazy breathing and no bad slow pan/push — prefer stillness, with subtle jitter as the only aliveness; cut at peak velocity with matched direction/speed (→ cut-catalog.md).

Part 3 — the seek-safe core (hard rules)

The frame is a paused GSAP timeline seeked frame-by-frame, so some “continuous” intents from a real-time engine can’t render — don’t name them. These are non-negotiable regardless of doctrine.

• No infinite / forever motion — “particles loop endlessly,” “logo rotates forever,” “marquee on repeat.” Any aliveness (the subtle jitter, a live SVG internal, a needed bounded ambient) is a finite tween over the hold, never repeat / yoyo.
• No randomness or wall-clock — no Math.random particle fields, no Date.now drift. Every render must be identical; name deterministic motion only (stagger and any variation derive from the element index).
• Entrances use fromTo — state the from-state explicitly so a seek to t=0 lands the element correctly; never rely on a CSS-hidden start (it renders visible before the tween claims it, and flickers under seek).
• No CSS transition / @keyframes for motion — CSS animation runs on the browser clock, independent of the HF seek clock; it desyncs and flickers. Drive all motion inside the paused GSAP timeline.
• Entrance + sequential reveal only — no mid-video exit. The frame unmounts via the harness transition; that injected transition_in is the exit. Exit motion belongs only to the final frame. (Worker-built seam cuts in cut-catalog.md are within-frame, not the frame’s exit.)

Forbidden — the failure modes

Slideshow (the primary failure): everything dumped on screen in the first ~25%; content enters then freezes; nothing revealed on its VO cue. Fix with rule 2 (sequential reveal timed to the VO).

Cheap aliveness: circular breathing as “life”; a slow pan/push in the back half disrupting the eye; many elements floating independently as “motion.” Fix with rule 3 (stillness + subtle jitter only).

Bouncy: back.out / bounce.out / elastic.out as the default entrance; hand-keyed overshoot. Fix with rule 1 (power3 long-tail; overshoot only when explicitly playful).

Always: no repeat / yoyo; no Math.random / Date.now; no all-elements-entering-simultaneously (sequence or stagger).

Naming motion in a shot — example

Scene 1 (0.0–1.0s): solid field; hero headline enters via per-word staggered reveal (dynamic-content-sequencing) on a smooth long-tail settle (power3); slow push on the root (multi-phase-camera) holds steady — no back-half re-push. Scene 2 (1.0–3.0s): as the VO names each capability, five feature icons reveal sequentially via cluster→outward expansion (center-outward-expansion), then a value-scaled counter (counting-dynamic-scale) ticks up beneath them — the back-half reveal, timed to the script, not dumped at t=0. Scene 3 (3.0–4.2s): hold on the result; keyword glow (asr-keyword-glow) lands on the payoff word as the VO says it; settles and holds still — at most subtle jitter (sine-wave-loop, low amplitude) keeps it alive; no breathing, no drift.

Name the move + its rule id (or cut-catalog.md for a seam cut) per scene; let the worker pick curves, ms, and stagger — defaulting to power3.