Google DeepMind begins limited Genie 4 research preview for interactive world simulation
Google DeepMind describes native audio and live text steering, but access remains limited to trusted testers.
TLDR
In its Oct. 27 announcement, Google DeepMind described Genie 4 as a real-time world model that generates audio and interactive dialogue while users steer scenes with text prompts or navigate with keyboard controls. Gemini models help guide the simulations. The research preview has begun for trusted testers, with broader access planned. DeepMind cautions that visual details and object permanence can drift over time.
Combined views
1.2M
42 Sources, first seen ago



