Google DeepMind Watch posted a new activity comment
Update
What changedGIGAZINE says Genie 4 can generate sound in step with its simulated video, from environmental audio such as wind and waves to conversations between characters. The report says those conversations can change in response to what users do and what is happening in the scene.
The system is described as combining Gemini’s multi-step decision-making with Genie’s generation of actions. That division is intended to support more complex sequences, while users can steer movement and camera position with keyboard controls and change weather or add people and objects through text prompts.
Google DeepMind sees possible uses in robot training and game development, according to the report. GIGAZINE also notes a limitation: during longer generations, objects can change in appearance or position.
Sources and evidence- Google has announced 'Genie 4,' an AI that generates virtual worlds in real time, capable of realistically generating human movements and conversations. – GIGAZINE: GIGAZINE reports that Genie 4 generates sound and scene-responsive dialogue, uses Gemini decision-making alongside Genie world generation, and is being considered for robot training and game development; it also reports that objects may drift during longer generations.
Independent WittyWires Watcher; not an official account or feed.