Recently, I came across the news that Claude can now connect with creative tools such as Blender and Autodesk Fusion. What caught my attention was not just the technical side, but the idea that AI is moving closer to the tools people actually use to create real work. Anthropic presents these connectors as a way to help with repetitive tasks, scripting, and faster iteration inside familiar software.
I do not have any real experience with Blender, but my daughter enjoys using it, so I decided to see what Anthropic and Blender had actually put together. At first, the results were modest. The models I generated felt very basic, built from simple geometry like cubes and spheres, and they did not look especially impressive.
Then I saw the news about Microsoft’s TRELLIS.2, a newer 3D generation model with a much stronger focus on high-fidelity image-to-3D output. Microsoft’s project page and Hugging Face model card describe TRELLIS.2 as a 4B-parameter system that can generate textured 3D assets and represent geometry and appearance more effectively than older approaches.
I could not find the newest model on Hugging Face, but the older community TRELLIS Space still performed surprisingly well. It handled my daughter’s favorite character, Hatsune Miku, exceptionally well, even if the result was not perfect. The Hugging Face Space is set up to turn images into a 3D asset and export a GLB file, which made it easy to test quickly.
I uploaded the model, rendered a short animation, and showed it to my daughter. She was not impressed. In her view, AI is still “slop” and she was not ready to be convinced otherwise.
Still, I left the experiment with a good feeling. Tools like Claude and TRELLIS.2 show that creative AI is becoming more practical, more integrated, and much more capable than it was only a short time ago. Maybe her opinion will change later. Maybe mine will, too. And when it does, I will be ready to try these tools again in my own work.