The critic stays independent: a skill kit for AI commercials that survive review

What it does

Motion Video Kit is a Claude Code skill (usable as plain context with any LLM) for producing launch-style commercials for real businesses using AI-generated footage, HTML/GSAP motion, and selective Three.js. It’s essentially a curated body of knowledge — a workflow, reference docs, and measurement scripts — distilled from studying 28 professional SaaS launch films and building three full sample commercials through many rounds of critique.

The interesting bit

The core idea is the “Gauntlet loop”: the builder is never the judge. Fresh critics review the actual render, verify findings item by item, and everything gets logged in a ledger. That separation-of-powers discipline, borrowed from how real studios work, is what most AI video workflows skip — and it’s why their output usually looks almost right.

Key highlights

• Motion grammar distilled into six rules plus a catalog of 16 reusable mechanisms, with pacing numbers.
• Measurable quality checks: frozen-frame detection, loudness (via ffmpeg’s ebur128 filter), contrast, and brand colour.
• Audio rules covering back-timing a real track so its ending lands on the end card, plus SFX mixing targets.
• Three.js patterns for deterministic, seekable scenes — exploded layers, projected callouts, photos pinned in 3D context.
• A business playbook with verticals, price anchors, a pilot-offer template, and honesty rules (never pass AI imagery off as a real customer).

Caveats

• Rendering works best with HyperFrames; the principles are renderer-agnostic but the tooling leans on it.
• Scripts require ffmpeg/ffprobe, and the component lab template needs Node 22+.
• Market prices are a September 2026 snapshot of public listings — the README itself says to re-verify before quoting.

Verdict

Worth a look if you’re trying to sell AI-assisted video work to actual clients and want a review process with teeth. If you just want a text-to-video prompt, this is the opposite — it’s the boring, rigorous part that makes output client-ready.