Skip to content

About

106 named video effects, each shown as a short clip. Tick the ones you want and copy one prompt for Claude Code or any AI video tool. English or 中文.

Topics

Resources

Contributing

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

The Video Effects Library home page: a row of playing clips and one shelf per category

Video Effects Library

106 named video effects, each shown as a short playing clip.
Tick the ones you want and copy one prompt for Claude Code or any AI video tool.

Open the library · 中文说明 · Use it · What's inside · Add a category · Rebuild the clips

MIT licence No install needed Node 20 or newer to rebuild


Why

People make videos with AI now, but most of us can't name the effect we want. "Make it cinematic" gets a random result. "Dolly zoom: the camera pulls back from 4.6 m to 15 m while the lens zooms in, over 3.4 seconds" gets the shot.

This library puts the real name on each effect, shows it moving, and gives you the exact words to ask for it.

What's inside

Category Effects What it covers
Camera moves 16 Push in, orbit, dolly zoom, whip pan, FPV fly-through and more, all filmed in the same 3D set so you can compare them
Motion graphics 16 Pop in, overshoot, stagger, mask reveal, count-up, motion trail and more
Design styles 46 Bauhaus, Swiss style, Risograph, Glassmorphism, Liquid metal, FUI, Technical drawing, Storybook illustration and more, each drawn and moving in its own style
Text animation 28 Typewriter, scramble, highlighter, karaoke captions, 3D extruded text and more

Every effect has:

  • a 9:16 clip made in code (no stock footage, no generated pictures)
  • its real name, backed by source links in the data file
  • How to ask an AI: one plain sentence
  • a motion spec: the exact move, distance, timing and easing
  • a prompt you can paste into Claude Code, Runway, Kling, Veo and other tools

The design styles page with filters for TikTok looks, art movements, print, screen and UI, 3D, retro and drawn styles One effect opened: the big clip, how to ask an AI, and the motion spec

Use it

Online: open video-effects-library.vercel.app.

On your own computer: no install and no server. Download the repo and open one file:

git clone https://github.com/ameng-nocode/video-effects-library.git

Then open site/index.html in your browser.

  1. Watch. Every clip plays. Use the tabs and filters to find a category or a type of effect.
  2. Tick. Tick the box under each effect you want. The Get prompt button counts them.
  3. Order and note. Open Get prompt, move effects up or down, and add a note to each ("for the opening").
  4. Copy. Press Copy prompt and paste it into your AI tool.

The copied prompt looks like this:

Use these effects in this order:

1. Push in (camera move)
Push in: the camera moves straight toward the subject, from 11 m to 6.4 m away, over 3.4 seconds with a slow start and a slow end. Near objects slide out of frame faster than the background.
My note: for the opening

2. Dolly zoom (camera move)
Dolly zoom: the camera pulls back from 4.6 m to 15 m while the lens zooms in at the same rate, so the subject stays the same size and the background grows behind it, over 3.4 seconds with a slow start and a slow end.

The top bar has an English / 中文 switch and a light / dark switch. Your picks, notes, language and theme are saved in your own browser.

The Get prompt panel: three picked effects in order, a note on the first one, and the prompt ready to copy The home page in Chinese with dark mode on

How it works

data/<category>.json        names, lines, prompts and sources for each effect
previews/clips/...          one 9:16 clip per effect (720 px wide, about 5 s)
site/categories.json        which categories the site shows, with their filter groups
site/build.mjs              reads the above and writes site/library.js
site/index.html             the site (plain HTML, CSS and JavaScript)
scripts/                    the code that builds every clip

The site reads only site/library.js, so it opens straight from disk. After you change any data, run:

node site/build.mjs

Add a category

A new category needs three things and no site code:

  1. A data file data/<category>.json: a list of effects, each with id, no, name, line, prompt, also_called ([{ "name", "where" }]) and sources ([{ "name", "url" }]).
  2. Clips at previews/clips/<category>/<id>.mp4: 9:16, 720 px wide, about 5 seconds, with the effect name and one line at the bottom like the others.
  3. One entry in site/categories.json with the title, one line, the data path, the clip folder and its filter groups. Add the Chinese title and filter names to site/zh.json.

Then run node site/build.mjs. "Transitions" is already in the list and shows as Soon until its data file exists.

Rebuild the clips

You only need this to change a clip or make new ones. You need Node 20 or newer, ffmpeg and the HyperFrames renderer (run through npx, nothing to install by hand).

Each category has a scene builder and a render script:

Category Build the scenes Render the clips
Camera moves node scripts/build-set3d.mjs bash scripts/build-camera-render.sh <id>
Motion graphics node scripts/build-mg.mjs bash scripts/build-mg-render.sh <id>
Design styles node scripts/build-ds.mjs bash scripts/build-ds-render.sh <id>
Text animation node scripts/build-text.mjs bash scripts/build-text-render.sh <id>

Pass one or more effect ids, for example push-in orbit. Scenes use Three.js for 3D and GSAP on one paused timeline, so every frame renders the same way every time.

Where the names come from

No name is made up. Each one is the word an AI video tool, an animation library or a design reference uses for that effect. Every effect lists its sources with links in data/<category>.json.

Camera move names come from the official guides of Runway, Google Veo, Kling, MiniMax Hailuo, Higgsfield, Luma and OpenAI Sora. Motion graphic and text names come from GSAP, Motion, Animate.css, After Effects and the HyperFrames registry. Design style names come from design history references and the style words people use in AI image and video tools.

The names belong to the tools and communities that use them; this project only links to them. The clips are original and made in code for this library.

Credits

  • HyperFrames (Apache-2.0) renders the clips.
  • Three.js (MIT) for the 3D scenes. The 3D text uses the Helvetiker typeface that ships with Three.js.
  • GSAP for timing and easing inside the scenes.
  • Geist and Geist Mono (SIL Open Font License 1.1) for the labels in the clips.
  • The site loads Plus Jakarta Sans and JetBrains Mono from Google Fonts (SIL Open Font License 1.1).

Contributing

New effects, better clips and new categories are welcome. See CONTRIBUTING.md.

Licence

MIT. The code and the clips are free to use, change and share.

About

106 named video effects, each shown as a short clip. Tick the ones you want and copy one prompt for Claude Code or any AI video tool. English or 中文.

Topics

Resources

Contributing

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages