Hands-on with GPT-6 Astra 3D: exploded views, rigging, and real-time motion capture

Exploring GPT-6 Astra's 3D Modeling: Exploded Views, Rigging, Motion Capture, and Video Generation

The author ran three experiments with GPT-6 Astra: first, an autonomous Blender build of a Touhou Project shrine with an exploded-view assembly animation; second, a browser-based first-person walkthrough with collision detection. The standout test was a full Psyduck character pipeline—modeling, rigging, and a real-time web mocap app that mirrors webcam movement. A final test used 3D modeling to drive AI video generation for a porcelain-firing explainer. Architecture and environment modeling worked well; character modeling still needs heavy manual tweaking, but overall delivery time and cost dropped by one to two orders of magnitude.