3Dviz Pro Max

3Dviz Pro Max is a loadable agent skill for creative 3D visualization. The repo viettranx/3dviz-pro-max is MIT, created 2026-09-10, language JavaScript. On 2026-09-13 GitHub listed 205 stars and 30 forks. Last push 2026-09-11. Homepage 3dviz.dev rebuilds counters from the repo at build time. It is a skill pack plus examples, not a hosted renderer.

Compare Generate Image if you only needed stills, Scientific Schematics if you needed 2D publication diagrams, or Claude Code if you needed the host that loads SKILL.md.

Key Features

• Ten-step workflow in skills/3dviz-pro-max/SKILL.md: intent, object reasoning, representation, construction routing, quality targets, first view, behavior, run-and-inspect, refine, honest report.
• Catalog on 2026-09-10: 223 recipes, 440 knowledge records, 22 kit blueprints, 37 style profiles, 37 runnable example studies, 21 GPU showcase frames.
• Hosts: Claude Code plugin (/plugin marketplace add viettranx/3dviz-pro-max) or a copied skill folder. Codex uses the same folder. Copy the whole folder, never SKILL.md alone.
• Capture helper: scripts/capture.py drives named views after you install Playwright. Without it, capture exits 3 and the agent must say no frame was observed.
• Blender optional: 4.2 LTS or newer is recommended. Without it, kits top out at T2. T4 ships nowhere.

Limitation: 205 stars is heat, not an audit. The before/after eval scored four with-skill rounds 19 to 25 out of 30 against a baseline of 21 at n=1. Marketplace install is untested on the compatibility page. Anatomy still needs verified geometry. Knowledge records are authoring guidance, not runtime observations.

Use Cases

• Claude Code / Codex sessions that should ship an explorable Three.js scene instead of a still or a video.
• Teaching and science viz using the 37 English studies (heart, gears, orbits) as starting points.
• People who should use anything2explainer instead if they wanted a Remotion explainer, not a 3D viewport.

Pricing

BodyParts3D anatomy meshes stay CC BY-SA 2.1 Japan. Fonts are OFL. The Harness Village GIF redistribution rights are still unconfirmed on the README.

Getting Started

1. Clone github.com/viettranx/3dviz-pro-max.
2. ln -s "$(pwd)/skills/3dviz-pro-max" ~/.claude/skills/3dviz-pro-max, or claude --plugin-dir /path/to/3dviz-pro-max.
3. Ask: "Find the installed 3dviz-pro-max skill. Tell me which local entrypoint you loaded. Do not generate the scene yet."
4. Then: "Use 3dviz-pro-max to build a small fantasy village I can explore."

First-party resource: SKILL.md and 3dviz.dev.

Frequently Asked Questions

Does it render for me?

No. The skill supplies workflow, data, and templates. Your agent creates, runs, and inspects the artifact.

Do I need a GPU?

Showcase frames used a real GPU. The Vite + Three.js examples run in a browser. Capture needs Chromium, not a training GPU.

Can I install only SKILL.md?

No. The README says the canonical skill is the whole folder.

Alternatives

• Generate Image: stills, not a viewport.
• Scientific Schematics: 2D publication diagrams.
• anything2explainer: Remotion explainer video.

Tips

1. Run the recognition prompt before you ask for a village.
2. Treat a failed capture as useful. Do not let the agent describe frames it did not see.
3. Read the license notes on BodyParts3D before you republish anatomy studies.

Conclusion

3Dviz Pro Max is the skill to load when a coding agent should build a scene you can orbit, not a slide. Start at 3dviz.dev or the repo.