Motion Design — Art Direction and Motion Craft
This skill turns adjectives into observable decisions. "Premium" is not gold gradients; it is restraint, spacing, controlled contrast, and considered timing. "Futuristic" is not random neon; it is a coherent grid, precise motion, and a limited palette.
Read references/MOTION-CRAFT.md for numbers, tables, and worked examples.
1. Start with a compact creative direction
Write these seven lines before any scene code:
Subject: one sentence
Audience: who and where they watch
Message: the one thing they must remember
Tone: 3 adjectives, each translated into a visual decision
Palette: 4–6 hex values with roles
Type: 1–2 families, weights, scale
Motion: speed, easing character, transition vocabulary
If the tone adjectives cannot be translated into decisions, the direction is not finished. Example translation:
| Adjective | Concrete decisions |
|---|---|
| Premium | Slow easing, wide tracking on display type, 1 accent colour, deep negative space, soft key light, no bounce |
| Energetic | Fast entrances (8–12 frames), 6–10% overshoot, cut-on-beat, high contrast, saturated accent |
| Trustworthy | Symmetric grid, mid-tone palette, slow reveals, stable camera, plain language, no gimmicks |
| Technical | Monospace accents, ruled lines, measurement ticks, precise linear motion, cool neutrals |
2. One system, then scenes
Define the system once and reuse it:
- Grid: 12 columns, 5–8% outer margin. Vertical/social formats need more margin because platform UI covers edges.
- Type scale: e.g. 96/64/40/28/20 px at 1080p. Never invent a size per scene.
- Spacing: 8 px base unit. Space communicates grouping more than lines or boxes do.
- Colour roles: background, surface, primary text, muted text, accent, semantic. One accent, used sparingly.
- Shape language: corner radius, stroke weight, and whether elements are flat, bevelled, or lit.
- Motion language: entrance style, exit style, transition vocabulary, and whether camera moves exist at all.
A style frame is cheaper than ten finished scenes. Render one frame that represents the whole piece and get it right first.
3. Composition rules that hold up
- One focal point per beat. If two things are equally bright, equally large, and equally moving, there is no hierarchy.
- Contrast is relative: if everything glows, nothing does.
- Negative space is a feature. A 30% empty frame reads as confidence.
- Align to something — a grid, an optical centre, or a deliberate diagonal. Near-alignment looks like a mistake.
- Keep the subject out of the outer 5% on social formats.
- For text over imagery, ensure a luminance difference of at least ~4.5:1 for body copy; add a scrim rather than a heavy shadow.
- Depth cues in 2D: scale, blur, opacity, overlap, parallax. Use one or two, not all four.
4. Timing
Duration communicates weight. Rough starting points at 30 FPS:
| Element | Entrance | Exit | Dwell |
|---|---|---|---|
| Short headline (1–3 words) | 12–18 f | 8–12 f | ≥ 24 f |
| Long headline | 18–24 f | 12 f | ≥ 45 f |
| Body line | 12–16 f | 8 f | ≥ 30 f |
| Logo reveal | 24–40 f | 12 f | ≥ 45 f |
| Chart | 30–45 f | 10 f | ≥ 60 f |
| Full-screen transition | 10–18 f | — | — |
| Camera move (3D) | 45–90 f | — | — |
Rules:
- Bigger distance or mass → longer duration.
- Groups move together, then stagger by 2–6 frames; do not stagger everything by the same amount.
- Leave a beat of stillness after an important reveal.
- If a viewer must read it, it must be on screen about twice as long as you think.
5. Easing
| Situation | Curve |
|---|---|
| Arriving and settling | ease-out (cubic or quart) |
| Leaving | ease-in |
| Moving between two rest states | ease-in-out |
| Mechanical / continuous / looping | linear |
| Physical snap | spring, damping 14–22, stiffness 90–180 |
| Camera | very soft ease-in-out, long duration |
Never ease every property identically. Position, opacity, and rotation can share a curve; scale usually wants slightly tighter timing. Avoid bounce unless the brand is playful — it dates quickly and fights legibility.
6. Transitions
Build a vocabulary of 3–4 and reuse them:
- Cut — default for pace. Cheapest and often best.
- Crossfade — 6–12 frames, for soft continuity.
- Mask wipe — directional, for a spatial relationship.
- Scale-through — pushing into an element, for a chapter change.
- Match cut — shape or motion continuity, the most satisfying and hardest.
Avoid: default zoom-blur on everything, star wipes, and transitions that contradict the story's direction of travel.
7. Colour
- Start from one dominant, one accent, and neutrals. Add a second accent only if it carries meaning.
- Check contrast at final size, not on a large monitor.
- Keep saturation controlled: highly saturated colours on large areas fatigue the eye.
- Dark backgrounds need slightly desaturated accents and lifted shadow detail; pure black plus pure white text is harsh.
- Use colour to encode meaning consistently (one colour for "before", another for "after"), not decoratively.
- Verify colour across tools: browser (sRGB), Blender (view transform), and encoder. A washed-out composite usually means mismatched view transforms, not a bad palette.
8. Typography
- Max two families: one display, one text (or one family with multiple weights).
- Display tracking: tight for large sans (-1% to -3%), wide for short all-caps labels (+8% to +20%).
- Body line height: 1.35–1.6. Display line height: 1.0–1.2.
- Line length: 45–75 characters. Headlines can be shorter.
- Avoid orphans and single-word last lines; rewrite the copy if needed.
- Numbers: tabular figures for counters and tables.
- Never stretch or condense a typeface to fit; change the layout or the size.
Right-to-left (Persian, Arabic, Hebrew)
- Set direction at the container level and use a shaping-capable font (e.g. Vazirmatn, IBM Plex Sans Arabic, Noto Naskh Arabic).
- Test shaping, joining, diacritics, and punctuation in a rendered frame, not in code.
- Mixed content: keep Latin brand names/URLs in their own spans with
dir="ltr", or the bidi algorithm will reorder punctuation. - Persian numerals vs Latin numerals: choose one system for the whole piece.
- Line height needs more room than Latin for the same optical size.
- Do not letter-space Arabic script — it breaks cursive joining.
9. Sound
- Decide the role of sound before choosing tracks: pace-setter, emotional bed, information (voice), or punctuation (SFX).
- Separate stems: voice, music, ambience, spot effects. Duck music 6–10 dB under voice.
- Accent the visual event, not the clip start.
- Fade every cut; a hard start or stop clicks.
- Leave headroom (-1 dBTP), avoid clipping, check mono compatibility for phone speakers.
- Silence is a tool. Dropping music for one beat before a reveal is stronger than adding a riser.
- License every track and record the source. Never assume "free download" means commercial use is allowed.
10. Accessibility and platform fit
- Captions for any spoken content; on-screen text also helps silent viewing, which is the majority on social.
- Do not encode meaning in colour alone.
- Avoid flashing above 3 Hz; no strobing.
- Keep essential text inside safe areas, at a size readable on a phone.
- Check the platform's current duration, aspect, codec, and bitrate requirements before delivery.
11. Review method
Watch the piece three times:
- Silent, at 50% size — does the hierarchy read? Is anything clipped?
- Full size with sound — does the rhythm work? Is the audio fighting the image?
- Frame by frame at boundaries — flashes, gaps, jumps, unreadable frames.
Then ask the three questions:
- What is the one thing the viewer remembers?
- Where does attention go first, and is that intended?
- What can be removed without losing the message?
Removal is usually the highest-value revision.
12. Deliverable for the design phase
brief.md— subject, audience, message, constraints, assumptions.style-guide.md— palette with roles and hex, type scale, spacing, motion language.storyboard.md— one row/block per scene: time, purpose, composition, copy, motion, transition, sound, tool.- 1–3 style frames — the visual target, approved before mass production.