Use huashu-art-motion to build code-driven art animations

Use this art animation skill when you want a coding agent to construct scenes, motion and transitions rather than only return a finished image. The repository README describes reference breakdown, art-style design, narration-driven visual tracks and QA. It is a repository of instructions, scripts and examples—not a hosted production SaaS.

If you are searching for a Claude Code animation skill or a Codex animation skill, this project uses the generic coding agent install route. It is not presented here as an official Anthropic or OpenAI package, and we have not verified agent-specific compatibility. Your coding agent and any external model services retain their own access and cost requirements.

The upstream installation command is:

npx skills add alchaincyf/huashu-art-motion

Inspect the repository and the installation scope before running it. Read SKILL.md after installation to select the task-specific references. The documented rendering prerequisites are uv, ffmpeg and Playwright Chromium. Browser installation is a separate dependency step; installing the skill alone does not establish that the render environment works.

The README illustrates rendering with uv run --with playwright .... We source-checked that example without executing it. Inspect each selected script for additional dependencies and file paths. Upstream setup.

• 35 style recipes and scene implementations: starting parameters, visual motifs, transitions and documented limitations.
• 9 explainer grammar cards: eight include demo films, parameterized clips and JSON examples; the ninth is presenter-led reference material.
• Full-narration reference code: mixed-style, whiteboard, Vox-inspired and presenter examples are snapshots with additional asset prerequisites, not universally ready-to-render projects.
• Long-scroll skeleton: a character moves across connected worlds; the repository includes three sample world segments.
• breakdown.py: analyzes a reference video into cuts, rhythm and motion cues.
• qa.py: checks numerical signals such as stability, rendering efficiency, motion, smoothness and framing cues. Numerical QA does not replace watching the output.
• Animation libraries and audio helpers: drawing, cameras, layout, charts, transitions and code-based soundtrack examples.

The repository documents these nine approaches:

1. Kurzgesagt-inspired: flat shapes and changes of scale.
2. Vox-inspired: information collage, cutouts and annotation.
3. Whiteboard: progressive line drawing for relationships and narration.
4. 3Blue1Brown-inspired (3b1b): transform visual objects to explain an idea.
5. Storytime: character reactions and comic timing.
6. Kinetic type: words and emphasis as motion.
7. Keynote UI: presentation cards, interface elements and large numbers.
8. Finance charts: introduce axes, data and focused annotations.
9. Presenter-led explainer: reference implementation requiring your own character assets.

A search for a whiteboard animation skill, 3b1b animation skill or Kurzgesagt animation skill can lead to these repository-documented grammars. The names describe inspiration, not official packages, integrations or affiliations with those creators. Eight runnable demo grammars is not nine ready-to-render projects. Grammar and demo inventory.

1. Review the upstream repository, then install the skill using its documented command.
2. Inspect SKILL.md and choose your task: reference breakdown, art scene, explainer segment or long-scroll composition.
3. Choose one style recipe or explainer grammar; read its limitations and required assets.
4. Inspect the script dependencies, rendering entrypoint and output directories. Check uv, ffmpeg and Playwright Chromium before starting.
5. Use synthetic inputs or assets you have permission to use in a disposable project.
6. Render one still or short sample segment before attempting a full film. The README's still example uses render.py --solo 09_postimp --stills 0.3 within its documented uv/Playwright command.
7. Inspect the actual output for legibility, framing, movement and missing assets.
8. Run the documented qa.py checks and review the film visually. The upstream workflow also calls for independent review.
9. Only after this small check, decide whether the setup is appropriate for personal or customer material.

This is a suggested source-backed evaluation sequence, not a successful render performed by AgentSkillsHub. Workflow source.

Code and documentation are MIT-licensed. The repository separately describes exceptions:

• Stroke derivative data in the SpaceX whiteboard reference retains the Arphic Public License.
• Bundled fonts retain their respective SIL OFL licenses; inspect the font license inventory and any redistribution conditions.
• Huashu character artwork and hero frames, including the same character in overview images and demo videos, are provided for demonstration. The repository MIT license does not grant unrelated reuse of those character assets.

Do not treat the root MIT label as permission to reuse every bundled visual. Supply your own appropriately licensed character material for a new presenter-led film. We link to upstream demos rather than republish restricted artwork. README exceptions · MIT license · Font license inventory.

The skill can direct a coding agent to execute local scripts, use ffmpeg and Playwright, read input files and write rendered assets. Review the chosen scripts and limit the project to copies of the files needed for the task.

Some character workflows use image-model-generated frames followed by code compositing. Depending on your agent and setup, that can send material to an external service and incur model charges. The repository does not establish a completely offline workflow or free model usage. Confirm the provider, file access and spending before generating assets.

If a render fails, first check the required browser, ffmpeg availability, expected paths and missing character/audio assets. An installation success or a chat response is not evidence that a usable video was produced.