Meshy announced a new agent capability.
You’re probably familiar with Meshy, one of the leading AI services that is capable of automatically generating 3D printable models. The service can accept either a text prompt, an image, or both and then within a few seconds create a representative 3D model.
Meshy competes with a few other similar services, such as Tripo. At this early stage of the game, it is important for these companies to capture as much of the market share as possible. But to do that, they have to offer something different than their competition, and that is precisely what Meshy is doing with this announcement.
The new feature is “agentic” processing. Some advanced AI users will instantly understand what this means, but let me explain for those that don’t.
The simplest way to interact with an AI chatbot is to enter a prompt and receive the result. It’s a kind of “transactional” interface: you send something in, you get something out. Then it’s up to you to do something with that result. For example, you might ask the chatbot to create the text for an important email, but then it is your responsibility to copy and paste that text into your email client. You’re doing the work; the AI is simply giving you things to use when you do that work.
Agentic AI is quite different, but it all starts with a prompt. When the AI produces the output, it is then able to actually do things with that result. In our example above, perhaps an agentic AI could call up the email app, create the email, and send it — all while you watch. The key here is that the AI has “agency”: it can do things, not just tell you things.
This is what Meshy has just done. They’ve given their AI agency.
Thus, the Meshy workflow becomes more complex and capable. You have a “conversation” with Meshy, and that allows you to iterate through a design. For example, you might generate a 3D model and then ask, “Hey, make those fingers much fatter!” And then Meshy would do so.
Meshy writes:
“Meshy 3D Agent Beta treats 3D creation not as a sequence of isolated technical steps, but as a guided, iterative process. Creators can describe their ideas in natural language, visually review results, and continue developing them with specific adjustments.”
But what about taking action? That’s not entirely clear. Meshy said:
“The new feature brings key steps of 3D production into an agent-based workflow, from the initial idea to an editable 3D asset.”
They do not mention any specific tools that the agent can use, but it is clear that you would get a 3D file at the end.
It seems to me that the Meshy 3D Agent could use a bit more agency. For example, perhaps it could integrate directly into slicing software or CAD tools. These may actually be possible, if Meshy hasn’t mentioned that capability. But it is implied by their use of the word “agent.”
Nevertheless, this is a step forward, since up to now text/image to 3D AI has been very transactional: try one prompt; try a different prompt if it didn’t work; repeat.
Now at least you can have a decent conversation with your 3D AI.
Via PRNewswire and Meshy
