AI NewsOpen sourceAnnouncement
viettranx released 3dviz-pro-max, an agent skill that turns an idea into a Three.js or Blender scene
3dviz-pro-max is an MIT-licensed Claude Code and Codex skill that ships 223 recipes and 22 kits for turning a description into a working Three.js or Blender scene, and the repository has 545 stars.
Image: GitHub
Why it mattersA marketer, product manager or founder describing a scene in plain words can now get a Three.js file they open in a browser, which pulls 3D work out of the specialist queue and back into a chat window.
Building a 3D scene for a webpage used to mean opening Blender or writing Three.js by hand. viettranx published 3dviz-pro-max, an MIT-licensed skill for Claude Code and Codex that takes a description in a chat window and returns a scene you can open in a browser.
What the repository ships
The description on the repository says it is an "agent skill for creative 3D visualization: turn an idea into a Three.js/Blender scene worth exploring". viettranx names the contents: 223 recipes, 440 knowledge records, 22 proved kits and 37 runnable studies. A recipe is a scene pattern the agent can pick from, and a kit is a proven scene the agent can copy from. The stack is Three.js and Blender, with glTF as the output format, so the finished scene runs anywhere a browser does.
The repository is licensed MIT. It was created on 10 September 2026 and last pushed on 11 September 2026, so the code has settled and the stars are still moving. As of the fetch for this item, the count sat at 545.
What the skill actually does
A Claude Code or Codex skill runs inside the agent's chat window. You ask for a scene, the agent reads its recipes, picks the closest one, writes the Three.js or Blender code, and opens it. The prompt does the specialist's job of translating "an animated hero for a launch page, product floating with soft shadows on a plain background" into the scene graph, camera, and lighting that produce it. That is the whole point of shipping 223 recipes rather than one general prompt: the agent has a menu of scenes that already work, and the description picks one from the menu.
Who it is for
3D on the web used to require a specialist. A designer would draft the scene, a developer would rebuild it in Three.js, and a marketing team would see the result three sprints later. A skill that reads a description and returns a working scene changes who can ask for a 3D hero: a marketer describing a launch, a product manager sketching an onboarding animation, a founder mocking up a demo. The scene is a plain glTF file, so a developer can pick it up, adjust materials, and land it in the site without a new toolchain.
The repository is early, the stars are recent, and the traction is what someone else's engineers have decided to try. There is no shipped case study yet. Try it locally first on a scene the team would have made anyway, see what comes back, and only then add it to a product page.
Source
Primary source: viettranx/3dviz-pro-max on GitHub.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.