Motion Graphics

Voice kinetic

A kinetic typography film that follows the voice, one beat per phrase

Picture shape: 16:9
Runs on Claude CodeCodexAntigravity
Watch on YouTube

Turn one spoken script into a 40-75 second horizontal kinetic typography film (1920x1080), where each key phrase hits the screen on the exact word the voice says.

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

The viewer hears a sentence and sees its key phrase land at that same moment. Giant words slam, fly, type, stretch, flip or count up, with RGB split, bursts and speed streaks, on a navy grid with a lime glow. It suits a proof reel or a manifesto: a claim, a product, and what it replaces. Everything is HTML, CSS and canvas animation, with no AI video.

How it works

Open your agent and type:

Làm video chữ chạy theo giọng giới thiệu sản phẩm của tôi, 45 giây

The agent writes the facts and the script first, then a storyboard of beats anchored to spoken words. The scripts generate the voice, align every word, plan the music and sound effects from the hits, mix to -14 LUFS and render with HyperFrames.

The story of the video

A film uses 8 to 13 beats chosen from 17 types, each starting on its first spoken word. Beats include a perspective floor, wire 3D solids, a spiral on the music drop, a camera tunnel, a card that flips to show the storyboard as code, and a phone playing your own demo clip beside a giant word. The synth drop lands on the name reveal, and the film ends on a call-to-action pill.

Voice and audio

  • The voice follows the language of your input, Vietnamese or English. Vietnamese uses a fast TV-commercial read from Gemini (voice Orus); English uses an ElevenLabs preset voice. A free local voice (VieNeu) is the fallback: slower and calmer.
  • A synth bed at 120-128 BPM, with sound effects planned from the hits.

Honesty rules built in

  • Every claim about how the video was made must be true, and a counted number is measured or labelled as an estimate.
  • No profanity, no third-party logos unless the video is about that product.

What you get

  • An MP4 at 1920x1080, 30 fps, -14 LUFS, with a cover image on the first frames.
  • Two review gates before the render, stamped by another model, subagent or person.
  • A filming tip if you want to play the film full screen on a laptop and record it with your phone.

What you need

  • Claude Code or Codex CLI, and Node.js.
  • For the Vietnamese default voice, a Gemini API key; for English, an ElevenLabs key. Without them, the local VieNeu voice works after a one-time setup.

Install and make your first film

Copy the skill folder into your skills folder:

~/.claude/skills/agentvid-motion-voice-kinetic-16x9/

For Codex, use ~/.agents/skills/agentvid-motion-voice-kinetic-16x9/. Then type a request like the one above.

Not a fit if you want

  • A product tour with real UI.
  • A long lecture.
  • A topic with nothing to prove.

Frequently asked questions

Can it use my own voice?

The look is built for a fast commercial-style AI read, not the channel owner's voice. An ElevenLabs cloned voice can be set in the project if you have one.

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