Google DeepMind Unveils Genie 4: Photorealistic Interactive 3D World Model with Real-Time Physics and Audio
Google DeepMind says Genie 4 integrates Gemini models for reasoning, with a research preview rolling out to trusted testers.
TLDR
Google DeepMind says Genie 4 generates explorable simulations with photorealistic visuals, native audio and real-time control through text prompts. It uses Gemini models to guide multi-step events and could help develop virtual training environments for robots and agents. Filmmakers and game developers are also testing it for creative work. A research preview has begun for trusted testers, though DeepMind says visual details and object permanence can drift over longer sessions or after complex edits.
Combined views
902K
3 Sources, first seen ago

