Communitygithub.com

by-motion

Video and animation skill. Motion for social media in Claude Design: animated Stories and graphic Reels — storyboard, timing, movement by preset, music and rhythm, checks. Use together with by-composition and by-presets.

by-motion とは?

by-motion is a Claude Code agent skill that video and animation skill. Motion for social media in Claude Design: animated Stories and graphic Reels — storyboard, timing, movement by preset, music and rhythm, checks. Use together with by-composition and by-presets.

対応✓Claude Code~Codex CLI~Cursor
npx skills add https://github.com/bysaneera/by-design/tree/HEAD/plugins/by-design/skills/by-motion

お気に入りのAIに質問する

このエージェントスキルを事前に読み込んだ状態で新しいチャットを開きます。

ドキュメント

by.motion — movement

This skill governs HOW and WHEN elements move. Every frame is built by by-composition (layers A–C, formats, fonts, logo, two languages, "Style from practice"), the frame's visual style — by the preset and layers in by-presets, the character of motion — by the "Character by preset" table below. Load both skills before working. If a rule here conflicts with layer A of by-composition — layer A wins.

Why motion — the main rule

Motion exists to get the point across fast in pictures: show text, a photo or an idea interactively so the viewer understands without reading long text. Motion explains, it doesn't decorate.

  • Start with the point. Before the storyboard write in one sentence what the viewer must understand and how motion will show it: "before → after" — a transformation; "works anywhere" — the product stays still, the world changes; "3 steps" — steps appear one by one; "2× growth" — a bar grows; "what it's made of" — ingredients assemble into the product. A technique from the library is chosen for this sentence, not for beauty.
  • Picture instead of words. If the thought can be shown (photo, number, diagram, product in use) — show it, cut the text to a caption. One thought per scene.
  • On topic. Motion is tied to meaning and object: tea pours, cream spreads, money is counted, a message arrives. An effect unrelated to the topic — remove.
  • Smooth. Motion is fluid and whole: eased curves (no jerks or linear stops), each move starts from the previous one, elements don't twitch or jump on the first frame, 30 fps without dropped frames. Sharpness (Bold, Street) is intentional and on the beat, not from being unfinished.
  • Test: watch the video once without sound — the point is clear within the first 2–3 seconds and matches the sentence from the first item. If not — change the technique, not the details.

Formats: animated Stories and graphic Reels (no filming: text, photos, illustrations, infographics). Both 9:16, 1080×1920. Built in Claude Design and rendered straight to MP4.

Workflow

  1. Memory and preset. As steps 0–0a of by-composition: "Style from practice", preset and layers. No Design System — "Starting a new brand".
  2. Question before building, if not clear from the request: duration (or how many Stories), whether there's a voice-over, whether the track should be baked into the file or added at publishing, whether it's for paid ads. In one message.
  3. Storyboard. First line — the point in one sentence and how motion will show it ("Why motion"). Scene table: # · time (start–end) · on screen · text · movement · music accent. For a Stories series — a row per Story. Choose 1–2 techniques (in videos over 20 s — up to 3, see "Combining techniques") from the "Technique library" (M1–M20) for the whole video and name them in the storyboard. Show the user before animating: editing a table is cheaper than a video. If there's no one to ask — go on and attach the storyboard to the summary.
  4. Key frames. Build each scene as a static artboard in the "everything in place" state by by-composition and by-presets and pass levels 1 and 1b. Motion won't fix a bad frame.
  5. Animation — per "Time", "Movement", "Building in Claude Design".
  6. Music and rhythm — per "Sound".
  7. Check — per "Check".
  8. Delivery — per "Delivery".

Time

  • Reading time. After text has fully appeared, it stays still for at least 1 s + 0.3 s per word (Japanese/Chinese: 1 s + ~0.1 s per character; Korean: per word group), minimum 1.5 s. Count by the longer language. A large number or 1–3 words — minimum 1.2 s.
  • Amount per frame — A1 limits for Stories (body text up to 25 words). Better less text and more scenes.
  • Hook in the first 1–2 s. The main phrase or number reads within the first second; don't start with an empty background, a logo or a long intro.
  • Ending. The last frame — a conclusion or CTA of up to 4 words and the logo (A6), held 1.5–2 s without movement.
  • Duration. Graphic Reels — usually 7–20 s, longer only if the text requires it. Stories — 5–15 s each, a series of 3–7.
  • Text rhythm: the headline appears before the body text; order of appearance = reading order (A2).

Movement

General rules

  • One movement per scene and no more than 2–3 elements moving at once. The main thing moves — the rest waits.
  • One set for the whole video: 2–3 enter/exit techniques, one set of easing curves and durations. Same set across a Stories series and a content pillar.
  • Curves: enter — ease-out, exit — ease-in, transition — ease-in-out. Linear only for continuous motion (ticker, slow drift).
  • Durations: element enter 300–600 ms, exit faster (200–400), scene change 300–500. Stagger between lines or elements 60–120 ms. The preset changes these numbers (table below).
  • Text appears by lines or words. By letters — only a hero of 1–3 words. While text is being read it doesn't move (a barely noticeable drift of background or photo is OK, no more than 2–3% per scene).
  • Numbers — a counter from 0 (or from "before") to the exact number in the text in 0.8–1.2 s, the final value held at least 1 s. A visualisation (ring, bars, line) draws in together with the counter and stops exactly at the value (A7).
  • Photos: slow push-in or pan (scale 100→105–108% per scene), a wipe mask on enter. The main person's face doesn't leave the frame and isn't covered by moving text (A12).
  • No flashes: no more than 3 sharp brightness changes per second, no flicker of contrasting colours — dangerous for some viewers.
  • Nothing "for liveliness": no meaningless motion (spinning icons, bouncing arrows, background particles) unless the preset calls for it.

Character by preset

The frame looks per the preset card in by-presets (colour, font, photo, graphics, layouts L); it moves per this table.

PresetMovementTempo
Neutral, Corporate, Fintechfade + rise 24–40 px, lines and charts draw in, number counterscalm, 400–600 ms
Editorialwipe masks, text line by line from under a rule, slow push-in on photosslow, 600–900 ms
Healthsoft fade, scale 98→100%, diagrams draw in as linescalm, 500–700 ms
Boldhard cuts, full-frame kinetic typography, word by word on the beatfast, 150–300 ms, cuts on the beat
Minimalopacity only, long pauses, one element at a timevery slow, 700–1000 ms
Playfulbounce and overshoot, stickers "pop", gentle wobblelively, 300–500 ms
Streetstop-motion (animation "on twos", 12 fps), jitter, layered collage; glitch — rarelyragged, on the beat
Lifestyleslow push-ins on photos, soft transitions through opacitycalm, 500–800 ms

Layers: Promo — the price or percentage "punches in" on the strong beat, the offer deadline stays longest; Cultural code — the ornament draws in as a line or unfolds from the centre; Holidays — one festive device per video, no fireworks in every scene; other layers move by the base preset. The brand and "Style from practice" outrank the table.

Technique library

Compiled from an analysis of ~100 motion-design references. A technique is a structure of movement; colour, font and photo come from the brand and preset. Examples explain the technique; their words, logos and characters are not copied.

Common to all references (we do the same): one technique per video, a palette of 2–3 colours, lots of empty space; the pattern "assemble → long pause on the finished frame → change"; text large and short; the end often loops into the start (seamless loop); videos 4–16 s.

Text

  • M1. Word by word → paragraph. Each word first fills the frame in turn, at the end they all settle into a normal headline; small photos may flash around. Hooks and manifestos. Presets: Bold, Editorial, Neutral.
  • M2. Word with pictures inside. Words move apart and photos or icons slot into the gaps, or pictures fly around a large word. Presets: Editorial, Lifestyle, Street.
  • M3. Word rotator. A list of words scrolls, one highlighted (brighter or on a plate), next to it a photo changes to match the word. For "who it's for / what we do". Presets: Corporate, Fintech, Editorial.
  • M4. Type beyond the frame. A word so large it's cropped, slides and reads in full; or a word in perspective. Presets: Bold, Street.
  • M5. Variable font and style changes. Letters change width, weight or style in place while lines stay readable. Variable font — only from the Design System; decorative style changes — per A11 (one decorative device per video). Presets: Bold, Street.
  • M6. Text on a path. Text runs along a spiral, loop or circle. Only a short repeated phrase; the main message separately as static text. Presets: Playful, Street, Cultural code (along an ornament line).
  • M7. Interface dialogue. Text appears as chat bubbles or notification cards. Presets: Fintech, Playful.

Shapes, line, drawing

  • M8. Shapes reveal the frame. Circles, blobs, plates grow and merge, revealing text or the next scene; shapes assemble into a logo mark or letter. Presets: any; in Minimal — one soft shape.
  • M9. A line draws in. Outline shapes fold and turn into a list or diagram; a signature or handwritten mark writes itself. Presets: Health, Minimal, Editorial, Cultural code (ornament).
  • M10. A flat illustration comes alive. A scene of simple shapes, objects hop in and sway, the colour field changes between scenes. Presets: Playful, Holidays, Food. Illustrations — per the by-presets illustration rule.
  • M11. Frame-by-frame "hand" animation. A drawing, cards or pixels change frame by frame like a flipbook, a sticker character changes pose. 8–12 fps. Presets: Street, Playful, Food.

Photo, collage, interface

  • M12. Window and split. One window frame on a field, the photo inside changes or expands to full frame, the headline stays; or the frame is split by a wipe into two photos. Presets: Editorial, Lifestyle, Real Estate, Food.
  • M13. Collage and stack. Photos, cut-outs, stickers and newspaper letters layer up, shuffle like cards, torn paper edges. Presets: Street, Editorial, Bold.
  • M14. Interface and numbers. App cards and phones slide in and scroll; rings and charts draw in, numbers count; portraits in circles orbit. All numbers — per A7. Presets: Fintech, Corporate, Health; product on a field — Lifestyle, Beauty, Food.

Series and voice

  • M15. Recap series ("the year in numbers"). A Stories series, one fact per Story: a teaser phrase line by line with a stagger ("Your most frequent colour this year…") → the answer large: a counting number or a photo grid. Background shapes (circles, stripes, arcs) flow from Story to Story and carry continuity; the background switches colour from the brand palette by section. Ending — a summary grid of all numbers. For year-end recaps, reports, anniversaries. All numbers — from the client's text (A7). Presets: Fintech, Corporate, Playful, Bold.
  • M16. Karaoke subtitles with a voice wave. The whole phrase on a flat field, the current word highlighted with a brand-colour plate in time with speech; below — the voice waveform. For quotes from podcasts, interviews, voice testimonials without video. Phrase within subtitle limits, phrase changes on speech pauses. Presets: any.
  • M17. Progress line. A thin line with a dot travels through the frame or flows from scene to scene; text and circle nodes appear where the dot passes. Links scenes instead of transitions. For steps, timelines, brand presentations. Presets: Minimal, Corporate, Fintech, Editorial.

Additional samples for earlier techniques: M13 — a cut-out person on changing colour fields; food: cut-out dishes and packaging on brand-colour tiles alternating with live cooking shots (Food layer); M14 — a static ad layout breaks into layers and reassembles; onboarding Stories with a question and buttons (where an Instagram sticker replaces the buttons — A10c).

Product and Beauty

  • M18. The product stays — the world changes. The product (in hand or on a podium) is still and always in the same place, while backgrounds change around it: views from a plane window, ingredient textures (rice, sand, water, fabric), day and night, flowers opening around a bottle. Hard cuts or on the beat, the headline stays. For "anywhere / for any situation / what it's made of". Presets: Beauty, Lifestyle, Food, Promo.
  • M19. The product assembles. Ingredients, capsules or fragments fly together into a jar or bottle; the lid comes down from above; ending — the product on a field of the same ingredients. Only from real photos or a 3D model of the product from the client (A9: packaging is not generated). Presets: Beauty, Health, Food.
  • M20. Texture and application. A macro of skin or texture: a swatch appears and spreads, a drop runs down; then the product name is hand-drawn (M9) or the frame opens like a book cover and shows application on the face. The face may be covered under the A12 exception if the video is about the product. Presets: Beauty, Health.

Also from beauty: words fly in an arc around the product (M6); a route on a map draws in and stops at the products (M9); a set of products on a string or ribbon rides through the frame.

Cultural code in motion

Status "observation" — refine on the first videos. Calm tempo (like Lifestyle), long pauses. Ornaments and symbols — per the "Cultural code" layer in by-presets (the motifs of the brand's market). Worked example for Central Asian / Kazakh motifs:

  • Living nature: steppe and grass in the wind, clouds over mountains, water, galloping horses — live video or a slow push-in and layered parallax over a photo (sky, mountains, foreground). The headline sits in a calm zone of sky or steppe.
  • Tea and the table: steam over a bowl, tea pouring in a stream, hands passing a bowl, a table filling from above (like M18: centre stays, things appear around).
  • Ornament: draws in as a line from the centre outwards (M9); unfolds symmetrically to 2 or 4 sides; runs as a border ribbon along the frame edge; a rosette rotates slowly (no more than one turn per video); works as a mask — a photo reveals in the shape of the ornament or a traditional architectural symbol. The ornament doesn't deform or "jump".
  • Transitions — dissolve, ornament mask or a through-line (M17); no hard cuts or glitch. For another market, replace the motifs with that culture's own (landscape, ritual, pattern) and keep the same rules.

Transitions

A transition is part of the technique, not a separate effect: it continues the movement and links the meaning of scenes. One type per video; common ones:

  • Through-element — the product, window, headline or line stays in place, everything around changes (M12, M17, M18). The cleanest, fits everything.
  • Shape wipe — a circle, blob or plate grows, covers the frame and opens the next (M8). Playful, Fintech, Health.
  • Colour-field swap — the background switches on a cut, the hero stays (M13, M15). Bold, Street, Playful.
  • Fly-in — the camera enters a window, letter, photo or circle and it becomes the next frame. Editorial, Lifestyle, Beauty.
  • Push and page-turn — cards and screens push each other, the frame opens like a book or a door (M14, M20). Fintech, Corporate, Beauty.
  • Cut on the beat — an instant change on the strong beat. Bold, Street, Promo.
  • Dissolve — through opacity, 400–800 ms. Minimal, Lifestyle, Cultural code.

Choosing

  • First the point ("Why motion"), then the technique for it.
  • The product and its story → M18–M20.
  • Recaps, anniversary, a series of facts in Stories → M15; voice without video → M16; steps and timeline → M17.
  • First "Style from practice" (it may hold the brand's technique), then techniques listing the brand's preset, then by content: numbers → M14; a list of audiences or services → M3; manifesto, hook → M1, M4; photo series → M12, M13; holiday, story → M10.
  • Combining techniques. Up to 7–20 s — no more than two techniques; in videos over 20 s — up to three, one each for hook, middle and ending, connected by a through-element (M17 line, product, headline or background shape). How to combine:
    • Lead + support: one technique sets the character, the second helps at points (M18 throughout + M20 at the end).
    • By parts: different techniques in different parts (M1 in the hook → M14 with numbers → ending).
    • Layered in one scene — only if one works with text and the other with the background (M2 over M13); two techniques on the same element — no.
    • In any combination — one set of curves, one preset tempo, one transition type: the video looks like one system, not a set of effects. The combination is chosen for the point; each technique shows its part of the thought.
    • In a content pillar — the same set of techniques episode after episode. A technique moves at the preset's tempo (table above): M1 in Bold is cut on the beat, in Editorial — slowly from under masks.
  • Reading time and zones apply to all techniques: decorative motion stops while text to be read is on screen.
  • New references (e.g. the user's own motion board) — analyse the same way: video frames at 7 points, technique → an existing M or a new number, with the user's OK.

Formats

Animated Stories

  • Zones — as in static Stories: nothing important in the top 250 px and bottom 340 px, including during motion (text doesn't fly in from under the progress bar).
  • Interactive (A10c): the sticker zone is free the whole time, motion doesn't cross it. The question appears and settles within the first 1.5 s, then the frame is calm.
  • Series: one set of movements and one headline position; each Story makes sense on its own (they get swiped). The first frame of each Story already reads.
  • A Story shorter than its reading time — no: duration = enter + reading time of all text + 0.5 s.

Graphic Reels

  • Zones — as for the Reels cover (A10b): text and faces in x 72–940, y 250–1500 for the whole video; ~420 px at the bottom for caption and audio, ~140 px on the right for buttons.
  • Structure: hook (1–2 s) → 3–6 scenes, one thought each → ending with the logo. One transition type between scenes for the whole video.
  • Cover — a separate static frame per A10b; best to take the hook scene in its final state.
  • The video is understandable without sound. With a voice-over — subtitles are mandatory: up to 2 lines, up to ~32 characters per line (Japanese/Chinese ~13–16 full-width characters, Korean ~16–18), 48–56 px, in y 1150–1450, the brand's body font, contrast per A8 (a calm background-colour plate is OK). Change by phrase, in sync with the voice; word highlight — only if the preset is lively (Bold, Playful, Street).

Two languages (A13)

  • Switch: frame in the primary language → the same frame in the second language. Reading time for each language separately, equal (by the longer one). Only the text changes (opacity transition 300 ms or a wipe); background, photo and shared elements (numbers, dates, logo) stay.
  • Both on screen: any mixing method from A13; the primary language appears first, the second follows 150–300 ms later.
  • Subtitles with two languages: the voice-over language as the main line, the second language a line below, smaller and in a tone; no more than 3 lines together.

Sound

The skill suggests a track and edits to the rhythm.

  • Track suggestion — 2–3 options: mood, genre, tempo (BPM), from which second to take the fragment. By preset: Corporate/Fintech — calm electronica 90–110 BPM; Bold/Street — hip-hop, phonk, drum 120–150; Minimal/Editorial — ambient, piano 60–90; Playful — indie pop, ukulele 100–120; Lifestyle — lo-fi, soul 80–100. A trending sound — only if the user sent it or it's in the Trend Library; don't invent track names — describe the character and where to look.
  • Rights. Better to add the track at publishing from Instagram's audio library. Business accounts and ads only have music without copyright restrictions — advise looking there. Bake into the file only a track licensed for commercial use (ask whether the client has one).
  • Editing to the beat. From the tempo: beat = 60 / BPM seconds (120 BPM → 0.5 s, a 4-beat bar → 2 s). Scene changes on the start of a bar, accents (number, price, logo) on the strong beat. Scene length is a whole number of bars, but not shorter than the reading time: if it doesn't fit — take the next bar, don't cut reading.
  • Mark tempo and bars in the storyboard; if the track will be added later — build to the chosen BPM so changes land on any track of that tempo.

Building in Claude Design

  • One 1080×1920 artboard per video or per Story; frame and margins from by-composition (--safe-top: 250px; --safe-bottom: 340px for Stories).
  • All timings in one place in the code: a scenes object (start, duration, text) and variables for durations and curves. So timing is checked in the code and changed in one place.
  • Animate transform and opacity (smooth), not width/top/left. All timings run off one timeline, not separate delays.
  • Fonts — per A11: the whole video = one series; the font is loaded before the animation starts (otherwise the first frames render in a fallback font).
  • Render straight to MP4 with Claude Design.

Check

Modes and independent reviewers — as in by-composition. In addition:

  • The point: the sentence "what the viewer must understand" is written in the storyboard; a reviewer without the brief retells it after one viewing without sound. Every movement is on topic; no jerks or jumps.
  • Key frames passed levels 1–3 of by-composition (viewer test — on still frames and on the video).
  • Timing from the code: for every text, still time ≥ 1 s + 0.3 s × words, or 0.1 s × characters for Japanese/Chinese (minimum 1.5 s); calculate and list violations in the summary.
  • The hook reads within 2 s; the ending with the logo holds 1.5–2 s.
  • Zones throughout: check frames every 0.5 s — text doesn't enter service zones or the sticker zone, the face isn't covered.
  • One set of movements and curves; no more than 2–3 elements moving at once; the character matches the preset.
  • Counters and visualisations stop exactly at the number in the text.
  • No flashes more than 3 per second.
  • Scene changes land on the bar of the chosen tempo; the video is understandable without sound; subtitles are in sync with the voice.
  • The final MP4 watched in full at 375 px width (as on a phone): correct duration, the brand fonts on the first frames, no jerks or black frames at start and end, everything readable on first viewing.

Delivery

  • Video: MP4 (H.264), 1080×1920, 30 fps (Street with stop-motion — also 30, with 12 fps motion inside), sRGB. Build straight to MP4 in Claude Design. Stories — a separate file each.
  • Names — per by-composition "Delivery": brand_pillar_YYYY-MM-DD_reel.mp4, _story01.mp4, cover _cover.png.
  • In the summary: the storyboard with final timings; 2–3 track options (mood, BPM, where to look) and from which second; which stickers to add to Stories; what didn't pass the check and why. After publishing — a line in the publishing log in learned.md (format "graphic Reels" / "animated Stories").

Learning

Same as "Learning on the project" in by-composition: the user's edits on motion ("slower", "no bounce", "sharper cuts") are recorded in learned.md as a separate line "Motion: …" in the "Rules" section, with the user's OK. Reels statistics (retention, completion) are analysed per "What worked — by statistics"; a retention drop in the first 2 s is a signal to redo the hook.


Author: Sanira Baitassova. Part of the by skills collection.

関連スキル