name: motion-library description: The "قاموس الموشن" — a dictionary of 68 named motion-graphics effects for vertical talking-head reels (orbit, lean-in, backdrop-word, the-bill, place-in-grid, whip-pan, karaoke-captions, split-screen…), each with what it looks like, what it is good for, and how to make it. Use it whenever the user sends effect names (often as «بغيت هاد الحركات: orbit, lean-in, …», copied from the library page), asks which motion fits a moment («شنو الحركة اللي تصلح ل…», "what effect for the hook / the CTA / a number"), wants to browse the motions, or asks to add named motions to a reel built with Remotion, the /effects skill, web-astra-reels or webastra-showreel.
Motion library (قاموس الموشن)
68 named effects in 8 categories. The full list, with a Darija description, a "good for" hint and the way to build each one, is in references/effects.md. The same data is in effects.json (fields: id, category, source, description_ar, good_for_ar, effects_options, how). demo/motion-library.html is the visual page: a looping demo per effect, filters, and a "نسخ الأسماء" button that copies the picked names.
Categories
| id | name | count |
|---|---|---|
| back | ورا الراس (behind the head) | 13 |
| cam | حركة الكاميرا (camera) | 8 |
| text | النص والعناوين (text, titles) | 10 |
| ui | كروت وواجهات (cards, UI) | 13 |
| frame | تكوين الكادر (framing) | 8 |
| trans | الانتقالات (transitions) | 8 |
| look | اللون والإضاءة (colour, light) | 4 |
| acc | لمسات (accents) | 4 |
Two kinds of effect
/effects(45) — rendered on real footage by the effects skill (claude-video-effects). Face tracking and a person mask let them sit behind the head. One effect:python scripts/fx.py <id> <video> <out.mp4> --start S --end E --text "..." --opt key=value. Several: write a plan JSON and runfx.py --plan. Check stills first with--still.- classic (23) — editing basics (jump-cut, whip-pan, karaoke-captions, picture-in-picture, motion-path…). Build them in Remotion with the recipe in
references/effects.md.
When the user sends names
- Read each name's entry in
references/effects.md. An unknown name: say so and offer the closest one in the same category. - Map each effect to the spoken line it belongs to (word timings from Whisper). Land the effect on the word, not on the sentence start.
- Pick the route:
/effectsnames on the raw footage, classic names in the Remotion composition. A reel can mix both: render the/effectspass first, then use that video as the face layer in Remotion. - Render, look at stills of each effect, then the full video.
When the user asks what fits
Use the "good for" hint and these defaults:
- Hook (first 1–2 s): locked-on, lean-in, headline-box, backdrop-word, red-collage.
- Explaining steps: split-screen, lineup, todo, checklist, the-bill, motion-path.
- Numbers and proof: number-counter, liquid-glass-graph, profile-wall, over-the-shoulder-zoom.
- Portfolio: noise, orbit, carousel-spin, phone-fan, place-in-grid.
- Twist or serious moment: spotlight, slow-push, moody, reject-stamp.
- CTA (end): comment-box, messages, tap.
Give 2–3 options with one line each, then the one you recommend.
Rules that came from real reels
- While the screen is split, the captions sit on the seam between the two halves.
- Keep a motion on screen long enough to read it; don't cut it away on the next word.
- Don't stack more than one big effect at a time; between effects, let the face breathe.
- Arabic text: reveal whole words, never cut letters (they change shape).