Google DeepMind Releases Genie 4 With Real-Time Text Editing and Native Sound
Google DeepMind has released Genie 4, its most advanced world model for generating interactive 3D worlds, to Google AI Ultra subscribers. The model was previously available only as a research preview and now arrives as part of a major update to Project Genie.
The headline feature is real-time editing through text. Users can reshape a world while they are inside it, changing the weather, the characters or the storyline with written prompts. Genie 4 also produces native sound for the first time. The audio is generated in real time and adapts to what is happening in the scene, rather than being added afterwards.
The older Genie 3 model is not left behind. It now supports sessions of up to three minutes, 360-degree panoramas and chainable worlds, which let users link separate creations together.
The update also adds a new AI agent that helps build worlds from spoken descriptions, so users can describe what they want aloud instead of typing. A community gallery lets people share the worlds they have made and remix those of others.
On safety, all generated content carries SynthID watermarks, which mark it as AI-generated, and passes through safety checks. Access to Genie 4 is limited to the Ultra tier for now. Google AI Pro members can join a waitlist.
Continue in the app — vote & join in ➔