Animated Explainer

Under the hood

How an everyday thing works, shown from the inside

Picture shape: 9:16
Runs on Claude CodeCodexAntigravity

Explain how one everyday thing really works, such as contactless payment, a QR code, a touchscreen or noise-cancelling earbuds, in a 60-90 second vertical video (1080x1920) with a computed cross-section.

Important: this is a skill for terminal AI coding agents (Claude Code, Codex) that runs on your own computer. It is not a prompt for the ChatGPT website. Check the requirements below before buying.

Why this look

You tap your card and it pays. No cable, no button. The viewer hears that paradox, then sees the mechanism in a magnified cut-away or on a sensor grid: gold field lines bend, readouts change, numbers move because the picture moves. Everything that moves is computed from the scene, so the picture and the number cannot disagree.

The screen has three fixed tiers:

  • Top: karaoke captions, one word at a time, with blue selection boxes on your keywords.
  • Middle: the exhibit, one scene at a time, on a dark graphite stage that can fade to paper for interface and document scenes.
  • Bottom: a narrator mascot. The default is Volt, a small robot engineer with 12 poses. You can make your own character from a picture.

The story: 8 to 12 scenes

  1. Paradox: the ordinary thing that should not work.
  2. Question: what happens in under a second?
  3. Reframe: "not X, it is Y", with one concrete number.
  4. Imagine: an everyday analogy.
  5. Mechanism with numbers, step by step.
  6. Evidence: a real source (a person, patent, paper or standard) with a yellow highlight on the claim.
  7. "That is also why...": one or two consequences.
  8. Close, then a "send this to someone who still thinks..." share scene.

Scene types include the cut-away, a pressed touch layer, a sensor grid, an electrode lattice, phone mock-ups with a tapping glove, source cards, bars, a gauge and a big counter.

How it works

Open your agent and type a request like this:

Make an under-the-hood video: how does contactless payment work? Use real sources

The agent web-checks every claim into a facts file, writes the script and storyboard, generates the voice-over in one take and renders an MP4. Model values on screen keep a footnote marking them as illustrative. Two review gates must be stamped before posting.

What you get

  • A 9:16 MP4 at 1080x1920, mixed to -14 LUFS.
  • A quiet ambient music bed and sound effects.
  • A voice in Vietnamese (default) or English.
  • A cover image step for posting.
  • Each video ends with a style-reference credit line (Style: <channel>, or Phong cách: <channel> in Vietnamese), as the guide requires, and the script must be your own words.

What you need

  • Claude Code or Codex, able to run the skill's Node.js and Python scripts.
  • A Gemini key, an ElevenLabs key and a Soniox key (for word timings).
  • Vietnamese voice: installed once with the skill's setup-vieneu.py.

Install

~/.claude/skills/agentvid-motion-under-the-hood-9x16/

For Codex, use ~/.agents/skills/agentvid-motion-under-the-hood-9x16/.

Who it is for

Science and tech channels for curious adults, explaining one everyday device per video.

Not a fit if you want

  • News or product promos.
  • A topic with no mechanism or no source.

Frequently asked questions

Are the readouts real measurements?

No. They are computed from the scene's model and labelled as illustrative on screen. Facts, names and years come from real sources.

Can I use the videos commercially?

Videos you render are yours for any commercial use. You may not resell, redistribute or share the skill files.

More

Related looks