Spatial AI: Google Gemini Now Generates 3D Models and Simulations

The modality of text and flat images has been surpassed. On April 10, 2026, Google DeepMind announced a critical update: the Gemini AI assistant has received the feature to generate interactive 3D models and physical simulations from text prompts.

This is not just a "toy" for the Android ecosystem. On-the-fly generation of 3D assets alters the pipeline for the video game industry, industrial design (CAD), and virtual reality (XR). The neural network’s ability to not only render geometry but also assign physical properties to simulations (gravity, collisions) opens the doors for the mass production of synthetic data. Developers of physical robots (Embodied AI) no longer need to manually create training environments—Google's algorithms will be able to generate endless 3D training rooms from a prompt.

Source: Google / The Verge