For those into AI tools, there is now a very intriguing option to use Hitem3D in workflows.
Hitem3D is one of a very few AI models designed to generate 3D geometry, like Tripo or Meshy. To use them, you typically will have a text or image interface, where you present a prompt and/or 2D image to provide guidance. The AI then attempts to generate a 3D model.
I’ve used this type of tool for a couple of years now, and they are increasingly good. It’s now quite possible to generate fairly accurate 3D models simply based on a single input image, something that would have been completely impossible a few years ago.
So what’s new here, then?
The news is that someone has published an OpenClaw skill to invoke Hitem3D.
Hold on, what’s OpenClaw?
If you somehow haven’t heard yet, OpenClaw is the fastest-growing open-source tool in history. Introduced only in January, it is a layer on top of AI models that makes things happen.
Here’s a good way to explain it: if you use ChatGPT or similar tools, you are always cutting and pasting things in and out of the tool. Basically, you are doing the work, and ChatGPT is giving you guidance.
OpenClaw is entirely different. It acts as an agent, meaning it can do the work. If you ask it to do something, it actually does the work. It doesn’t just draft the text of your email, it drafts it AND SENDS IT. And that’s just the beginning — OpenClaw can, in theory, do almost anything you can physically do on your computer. Click links, read pages, make spreadsheets, edit files, install applications, etc.
Does it know how to do everything? Yes and no. Yes because it is possible to provide OpenClaw with a set of “instructions” on how to use an external tool. For example, you might have instructions to interface with a database or manage Google documents. These instructions are called “skills”.
Of course, there are significant security concerns, but it is entirely possible to set up OpenClaw to operate in a safe manner. However, that’s beyond the scope of this story.
As you might expect with such a tool, there are plenty of folks that are pushing the envelope on its use. Some have set up complex workflows involving multiple OpenClaw agents, each doing specialized work to accomplish a larger goal. For example, some literally create a software development “team” of agents to create large software systems in hours. It’s almost beyond belief what can be done.
Back to Hitem3D. The “skill” announced allows OpenClaw to invoke Hitem3D and generate 3D models from within an AI agent workflow. All you’ll need is an API key from Hitem3D, and your agent can then generate 3D models at once. Math Magic explains:
“The Skill encapsulates Hitem3D’s generation capabilities into a standardized execution flow. Within OpenClaw, agents first verify API credentials, then detect task types—such as single-image, multi-view, batch processing, or portrait generation—before confirming parameters including model version, resolution, output format, and generation mode.
Once configured, the agent submits the generation job via API, polls execution status, and returns downloadable results along with a structured parameter summary. In cases of failure, workflows may include retry guidance, such as adjusting resolution or input quality.”
It’s possible to generate results in a variety of formats, including GLB, OBJ, STL, FBX, and USDZ, so the output can be used both for visual applications and 3D printing. There are even 3D printing-specific sub-instructions:
“3D printing
If the user mentions 3D printing, printer, slicer, or STL, prefer:
format: 3 (STL)
request type: 1 for mesh-only when texture is irrelevant
optionally set higher face count for smoother geometry”
This means that 3D content can be produced that will generate 3D models in a much more rapid manner, perhaps even automatically. This is such a new capability that I am not sure of the applications that might emerge from this, but it’s heading in a very science fiction direction.
Via Hitem3D, ClawHub, and PRNewswire
