One of the more genuinely surprising ClawHub releases this month: Hitem3D, developed by Math Magic, is now available as a callable OpenClaw skill. It allows an agent to take an image as input and return a downloadable 3D model — in standard formats including GLB, OBJ, STL, FBX, and USDZ — as part of a fully automated workflow.
Image-to-3D generation has existed as a standalone tool for a while. What is new here is that it is now composable inside an agent pipeline. That changes what is possible.
What the Skill Does
Hitem3D wraps the Hitem3D v2.0 API into a structured OpenClaw skill. When triggered, the agent works through a defined sequence: it verifies API credentials, detects what kind of generation task is being requested, confirms the parameters (model version, resolution, output format, generation mode), submits the job via API, polls for completion, and returns the downloadable result alongside a structured parameter summary.
The skill supports four task types:
- Single-image generation — one photo in, one 3D model out
- Multi-view generation — uses several angles of the same subject for higher fidelity
- Batch processing — handles multiple images in a single job, useful for product catalogs or asset libraries
- Portrait generation — optimized for human subjects
A notable technical detail: Hitem3D v2.0 integrates geometry and texture generation within a unified pipeline rather than treating them as separate passes. The result is better surface consistency and fewer visible seams — one of the most common quality problems in image-to-3D workflows.
Output Formats and 3D Printing
The supported output formats cover most professional workflows: GLB and USDZ for web and AR/VR rendering, OBJ and FBX for game engines and 3D software, and STL for 3D printing. The STL outputs are designed to work with standard slicing software with minimal manual cleanup before printing — which is a meaningful claim, since raw AI-generated meshes often require significant repair before they are print-ready.
What This Unlocks as an Agent Skill
The interesting part is not the 3D generation itself — that technology exists. The interesting part is what becomes possible when it is composable inside an agent workflow.
Consider a product photography pipeline: an agent receives a product image, generates a 3D model via Hitem3D, converts it to USDZ for AR preview, and returns both the model and a rendered web-ready GLB — without a 3D artist touching the project. Or a manufacturing workflow where an agent takes reference photos of a physical object and outputs an STL file ready for prototyping review. Or a game asset pipeline where a batch of concept art images is converted to 3D mesh drafts for a technical artist to refine.
None of these workflows require a human to open a 3D tool, wait for a web interface, or manage file downloads manually. The agent handles the full cycle.
Context: A Significant ClawHub Week
The Hitem3D release coincides with OpenClaw v2026.3.22, which made ClawHub the default plugin and skill registry — replacing npm as the first-stop for skill installation. That structural change means ClawHub’s skill ecosystem now has more visibility and a cleaner installation path than it did a week ago. Skills like Hitem3D land in a registry that is genuinely easier to use than it was a month ago.
How to Install
The skill is available now on ClawHub. You will need a Hitem3D API key from hitem3d.ai to use it.
npx clawhub@latest install hitem3dFor anyone working in product design, manufacturing, game development, or AR/VR content creation, this is worth testing. The batch processing mode in particular makes it practical for production use rather than just experimentation.
Leave a Reply