From Ars Technica · · 1m 40s
DeepMind's Genie 4 adds audio and realistic humans to AI worlds
Google DeepMind releases an upgraded world model with better physics, native audio, and human speech, rolling out to paying users soon.
In brief
Google DeepMind releases an upgraded world model with better physics, native audio, and human speech, rolling out to paying users soon. DeepMind unveiled Genie 4, an improved world model generating realistic simulations with physics, audio and interactive dialogue, launching via paid API access. Originally reported by…
DeepMind upgrades its world model
Google DeepMind has released Genie 4, an updated version of its Genie 3 world model that improves on the original in almost every way.
What world models do
A world model creates a virtual world without pre-rendering or programming. You tell it what you want and it generates the experience frame by frame in real time.
Gemini steers the simulation
Genie 4 uses Gemini 3.8 Flash as a guiding agent to break down goals and evaluate each output frame, keeping the simulation on track through varying environments.
Physics and people look more real
Genie 4 handles aerodynamics and collisions more realistically. People in simulations move naturally and respond to environmental changes, looking less uncanny than before.
Audio and interactive speech arrive
Genie 4 is the first DeepMind world model with native audio, including ambient sounds and specified audio elements. It can create interactive dialogue that adapts as you change the simulation.
Pricing starts at one hundred dollars
Google will roll out Genie 4 via the Gemini API and AI Studio for developers. Consumers get access through AI Ultra plans starting at one hundred dollars per month.
Access rolling out in phases
DeepMind is working with a small group of trusted testers and plans to expand access soon with no specific timeline yet. A waitlist is available for early access.
What to remember
DeepMind unveiled Genie 4, an improved world model generating realistic simulations with physics, audio and interactive dialogue, launching via paid API access.
