agentleFS
alisadiq-aialisadiq-ai/favstash-ai-video-editing-skills/skills/3-styles-on-camera/motion-effects/SKILL.md

motion-effects

Skill75 starsChanged 39 days agoApache-2.0View on GitHub

Style for edit-video: a library of 33 named motion effects for talking-head reels (kinetic hook, gravity drop, timeline, title card, comment drop, orbit, frame break, vertigo, paper cut, rewind and more), built in one design system with words behind the speaker and the creator's own footage inside the UI. Use when the creator names an effect, asks for heavy motion graphics, or wants an effects listicle.

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Check the cutout over a flat colour before using it: hair, caps and glasses should hold.

    line 63
  2. Very specific:

    Run edit-video's edge check on the export.

    line 74

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Motion effects

A style on top of edit-video, which owns the cut, sound and review. The creator asks for an effect by name ("add a gravity drop on 'drop'") and you build it on the spoken word, in one consistent design system. The effect recipes hold the layers, timing, sound and a copy-paste prompt for each one.

What makes these look premium

Most AI edits look like stickers on a talking head. These don't, because of five habits:

  1. Depth. Big words sit behind the speaker and in front of the room. Cut the person out once and stack: camera, words, person, UI.
  2. One design system per reel. One dark colour taken from the room, one light colour, one accent. One serif, one heavy condensed sans, the system UI font for anything that imitates an app.
  3. Real content inside the UI. Frames of the creator's own take, their real posts, their real waveform, colours sampled from their frame. Never grey bars or fake text where real content exists.
  4. Native, small UI. App-like pills, fields and cards at phone scale, crisp shadows, one per beat, never covering the face.
  5. Restraint. A caption only for the hook and the call to action; everything else is the effect itself.

The design system (starting defaults)

ElementSetting
InkThe room's darkest colour, deepened (for a purple-lit room, #0a0f2c)
LightWarm cream #f4eee3 for big words and title cards; white for app UI
AccentOne colour (default #5b9eff), used for a single word or state per beat
SerifDidot or Bodoni 72 (system fonts on macOS); any high-contrast serif elsewhere
CondensedAvenir Next Condensed Heavy (macOS); any heavy condensed sans elsewhere
UI-apple-system / SF Pro, so app imitations read as native
CaptionSystem UI bold 46px, white, soft dark shadow, at chest height

Pick the ink from the actual footage; the rest of the table stays.

Layers and setup

Stack inside one camera wrapper so shakes, punch-ins and whips move everything together:

  1. the camera video
  2. the behind layer (big words, the post grid)
  3. the person cutout video
  4. effect overlays (grades, film, wipes), then the UI, then a checklist and captions

Make the cutout once per take:

npx hyperframes remove-background take.mp4 -o person.webm --quality best

Re-encode both the camera and the cutout with a keyframe every 15 frames before rendering (-g 15 -keyint_min 15; the cutout as VP9 with yuva420p). Sparse keyframes make HyperFrames seek badly and the two layers drift out of sync.

Check the cutout over a flat colour before using it: hair, caps and glasses should hold.

Placement and timing

  • Each effect lands on the word that names it (or the word it illustrates) and clears about 4 frames before the next one lands.
  • Behind-the-head words go in the empty band above the head and are sized to end before any side panel; let the head cover the bottom of the last line.
  • App UI sits in the top third (y 230–700) or at chest height (y 1330–1600), never over the mouth.
  • Text stays inside x 108–972 on a 1080-wide frame (what a 20:9 phone shows). Run edit-video's edge check on the export.
  • Optional checklist for listicles: a quiet panel at the top right (x 746–962) listing every effect, the current one in a light pill, done ones ticked.

Sound

One sound per visible event, on the frame it happens: an impact when a word lands, a short whoosh for a whip or a wipe, keys under typing, a click on a tap, a ding on a completed state. Keep them 4–8 dB under the voice, and let subtitle changes stay silent.

Review

Watch the export at 4 frames a second before sending: no clipped or stray letters, no text over the face, every effect on its word, the cutout aligned with the camera. Then run edit-video's first-second and edge checks.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.