| name | motion-design |
| description | Motion design creation — logo animation, animated ads, brand motion, kinetic graphics, promo videos. Trigger whenever the user says "make a motion", "motion design", "animate this", "make a video from my logo", "animated brand", "kinetic graphics", "promo video", or "ad video". Use for short-form branded motion; for narrative films with a script and multiple acts use film-setup instead. |
Motion Design
Guide the user through a full motion design flow. Be concise and direct. Speak the user's language.
Scope: short-form branded motion (5–15s). For a narrative film with a script, acts and voiceover, run film-setup instead — this skill deliberately skips scripting and casting.
STEP 0 — Flow type
classicMD — ads, brand promos, service presentations, logo reveals, atmospheric content.
highMD — sports, tech launches, music teasers, AI demos, fashion drops. Extreme camera speed, aggressive cuts, peak dynamics. Realistic people are replaced by silhouettes, chrome, or 3D abstract figures.
Obvious from the request? Proceed silently. Ambiguous? Ask once:
"Which fits better — Classic (smooth, elegant, cinematic) or Hyper/Kinetic (fast cuts, aggressive, CGI energy)?"
STEP 1 — Brief intake, ONE message
Ask everything at once. Never split across turns.
- Assets? — logo / product photo / reference, or none
- Duration — 5s teaser · 10s post · 15s promo
- Format — 16:9 · 9:16 · 1:1
- Mood — energetic, minimalist, luxury, technological, atmospheric, aggressive, cinematic
- Brand name + tagline
- Storyboard frames — 6 standard · 8 detailed · 9 maximum
STEP 2 — Assets
Has assets
Ask for the upload. Register it and proceed.
Hard rule — never let a model redraw a logo. Generative models approximate letterforms and fine detail. Use the logo as a reference for palette and mood, then composite the real file in post for any frame where it is a hero element. See brand-lock.
No assets
Generate a base visual. Preflight the cost first, then pick by need:
| Need | Model | Credits |
|---|
| Budget / previz | nano_banana | 1 |
| Text or typography in frame | gpt_image_2 2k medium | 3 |
| Default | nano_banana_pro 2k | 2 |
| Poster / crop-into | nano_banana_pro 4k | 4 |
2k costs the same as 1k — never generate at 1k.
Show the result. Ask: "Does this work, or would you like changes?" Iterate here — stills are 20× cheaper than clips.
STEP 3 — Storyboard
Generate one image containing all N panels in a grid — a single call, not N separate generations.
Use the approved asset as a reference. Prompt shape:
Storyboard sheet with [N] sequential panels in a grid, each labelled "Frame 1", "Frame 2"…
Panel 1: [scene]. Panel 2: [scene]. … Panel N: [logo lock].
Each panel: [camera angle], [motion state], [mood/lighting].
Style: [cinematic/kinetic]. Consistent palette. Thin borders. [aspect per panel].
classicMD — smooth compositions, elegant typography zones, cinematic lighting.
highMD — peak-action freeze frames, frozen splashes, shattered elements, material stretch, aggressive angles, neon contrast.
Present the sheet plus a summary:
Storyboard — [Brand]
🎬 Frame 1: … ✨ Frame 2: … 🏁 Frame N: logo lock
Mood: … Motion: … Ending: …
⛔ Approval gate
Do not generate video until the storyboard is approved. A storyboard costs ~2 credits; a clip costs 8–45. Iterate freely here.
STEP 4 — Video
Pick the tier the shot actually needs — do not default to the most expensive model:
| Need | Model | cr/sec |
|---|
| Abstract motion, logo reveal, no faces | kling3_0 pro | 1.75 |
| Best quality per credit | veo3_1 fast (8s clips) | 2.75 |
| Recurring recognisable face must hold | seedance_2_0 1080p | 9 |
Preflight before submitting:
higgsfield generate cost <model> --prompt x --duration 5 --resolution 1080p
Prompt construction:
classicMD — smooth motion design, [scene flow], elegant transitions, [mood] atmosphere, cinematic camera movement, [duration]s, brand reveal at end: [brand], [aspect]
highMD — high-intensity kinetic motion, [scene flow], extreme camera speed, aggressive match-cuts, peak-action freeze frames, [mood] CGI aesthetic, neon contrast, [duration]s, hard stop logo lock: [brand], [aspect]
highMD logo lock must be a static hold, built into the prompt: ~1s for 5s clips, ~2s for 10s, ~2–3s for 15s.
Pass the original asset as start_image when the user supplied one; otherwise the approved storyboard frame.
Operational traps: concurrency caps near 6 — submit in waves. Dark or moody prompts may be intercepted by a preset recommendation; pass declined_preset_id to proceed.
STEP 5 — Post, then review
Post is free. Composite the real logo, add typography, grade, and normalise loudness locally in ffmpeg/ImageMagick rather than regenerating.
Present and ask: "What do you think?"
- Love it — done
- Different edit — regenerate with adjusted prompt, same storyboard
- Different style — back to Step 1
- Another version — vary the prompt slightly
Rules
- All intake questions in one message
- Storyboard = one image, one call — never N separate generations
- No moodboard step — brief straight to storyboard
- highMD: no realistic humans — silhouettes, chrome, 3D abstract only
- classicMD logo may open, close, or both — ask if unspecified
- Match the user's language throughout
- Never render a logo generatively for a hero frame — composite the real file
- Disclose AI generation where footage depicts realistic people
- Check credits with
higgsfield account status before a batch; report cost when it is material
