Video Creation Skills for Claude Code, Tested: Explainers, Motion Graphics, 3D

Video is the fastest-growing skill category, and the technology story underneath is interesting: the best skills don't generate pixels — they write code that draws every frame. Here's the full collection, evaluated.

The video category grew from nothing to one of the deepest collections in the directory in a matter of weeks, and it splits cleanly into two philosophies.

Model-generated video — calling an image/video generation API for every scene — produces beautiful demos that fall apart on the second render. Code-drawn video — Remotion, manim, Three.js — writes programs that draw every frame, which means the output is debuggable, reproducible, and brandable. The leaderboard is dominated by the second philosophy, and the scores reflect it.

Here's the collection, top to bottom.

1. remotion-to-hyperframes — 9.7/10

Ports existing Remotion (React) compositions into the HyperFrames format.

Why: The #1 skill in the category is infrastructure. Every conversion step is verified, and the output is designed to be audited — portable, inspectable, reusable video programs. This is the rare tool that makes everything built on top of it better.

The one thing: You need to know what Remotion is first. If you're not already writing Remotion, start with create-video and come back.

2. create-video — 9.6/10

The flagship end-to-end skill: text prompt to final video via HeyGen's Video Agent.

Why: The one to install if you're starting. Full pipeline — script, visuals, timing, assembly — with verification at each stage, and task-routing that sends different video types to proven recipes.

The one thing: Defaults are opinionated. First output will look "create-video flavored" until you customize the style contract.

3. website-to-video — 9.5/10

Any website into a polished demo or explainer video — scrolling shots, feature callouts, transitions.

Why: The most immediately practical skill in the batch. Product videos are a constant need; this generates them from the live site with no footage. The pacing logic is genuinely smart.

The one thing: Output quality tracks the source site. A rough site gets a rough video.

4. music-to-video — 9.4/10

Music-synced visual videos — beat detection, tempo mapping, scene cutting driven by the music's structure.

Why: Audio-visual sync is where most AI video fails embarrassingly. This skill's verification loop checks that cuts land on beats — the difference between "synced" and "roughly near the beat."

The one thing: Style presets lean abstract/visualizer. Narrative music videos aren't its strength.

5. anything2explainer — 9.4/10

Topic in, narrated explainer video out — motion graphics, TTS voiceover, subtitles, chapter progress bar. Chinese or English.

Why: The most complete single-purpose pipeline evaluated. Every frame drawn in code with Remotion — nothing generated, nothing fake. "Kurzgesagt-style content, from a prompt."

The one thing: Needs a TTS API key configured upfront. Five minutes of setup, but it's not zero-config.

6. video-recap — 9.1/10

Turns existing video content into recaps and highlights.

Why: Recap workflows are structurally hard — they require understanding a source video well enough to compress it. The skill's analysis-before-editing order is what makes the output watchable instead of random clips.

The one thing: Recaps inherit the source's quality ceiling. Great source, great recap; mediocre source, recap of mediocre.

7. 3dviz-pro-max — 8.7/10

Expressive 3D scenes, explainers, and interactive models in the browser.

Why: Three.js from an agent is usually a spaghetti demo; this skill produces scenes with composition and intent. The interactive-model output is unique in the category — most video skills can't produce something you can rotate.

The one thing: 3D is compute-heavy. Complex scenes need a real GPU to develop against.

8. qisi-video-remix — 8.7/10

Creative adaptation from a reference video — readable script, character styling, storyboard, and independently usable segment prompts.

Why: The workflow is honest about its decomposition: it doesn't pretend to one-shot a full video from a reference. It produces the full pre-production artifact set, and each artifact is usable on its own.

The one thing: Output is pre-production material (scripts, storyboards, prompts), not final rendered video.

9. manim-composer — 8.4/10

Orchestrates manim — the math-animation engine behind 3Blue1Brown — for programmatic mathematical animation.

Why: Manim is powerful but notoriously fiddly. This wraps it with structured composition — scenes, timing, transitions — so your agent produces math explainers without the manim learning curve.

The one thing: The lowest score in the collection reflects the domain. Mathematical animation is hard; expect iteration.

The pattern

Across the whole category, three findings:

1. Code-drawn beats model-generated. Reproducible, brandable, artifact-free — every time.
2. Infrastructure is undervalued. The top of the leaderboard is pipeline tools (#1, the hyperframes family), not creative front-ends.
3. Verification is the differentiator. Skills that check their own output — cuts on beats, renders that complete — score a full tier above those that render and hope.

Full scorecards with written rationale are on the video creation collection page.

The ranked shortlist with all six dimension scores is at best video creation skills for Claude Code.