Google DeepMind unveils Genie 4 world model with native audio and two-minute scenes

Google DeepMind, the AI lab of Alphabet's Google, announced Genie 4 on Oct 27, a world model that generates interactive environments in real time together with their sound. The model is in a research preview for a limited group of trusted testers.

A world model builds a scene frame by frame as the user moves through it with a keyboard, instead of playing back a finished video. Google says the technology can provide training environments for robots and AI agents and creative tools for gaming, filmmaking and manufacturing.

What is new

• Genie 4 keeps a scene consistent for up to two minutes, double the limit of Genie 3.
• It generates audio on the fly, including ambient sound, music and character dialogue.
• Users can change a running scene with text prompts, such as moving the camera, changing the weather or adding a character.
• Google's Gemini models plan and check the simulation in the background, with Gemini 3.8 Flash acting as the guiding agent.
• Every video and audio clip carries SynthID, Google's hidden watermark for AI-generated content.

In one robotics experiment, a Gemini model broke the task of playing checkers into steps, and Genie 4 simulated Apptronik's Apollo humanoid robot carrying them out with no real robot involved.

Google says the model still has gaps. Details and objects can drift in longer simulations, and accurate audio and legible in-world text are not yet solved. DeepMind engineer Diego Rivas told reporters it is not yet clear what the end product will be.

Who is exposed

Alphabet owns DeepMind and plans to offer Genie 4 through its paid Google AI Ultra subscription and its developer tools.

Video game companies face Genie as a rival way to build interactive worlds. When Google began rolling out Project Genie, a prototype built on Genie 3, Unity Software shares closed down 24% on Jan 30, their worst one-day drop since 2022, Bloomberg reported. Take-Two Interactive and Roblox shares also fell that day, Reuters reported.

The world-model field is consolidating. TechCrunch reported that AMD is acquiring Fei-Fei Li's World Labs, Amazon is in talks to buy Decart, and Rivian acquired Third Dimension for self-driving simulation.

What comes next

• Google says Genie 4 will reach Google AI Ultra subscribers in Project Genie, and early-access partners of its Gemini developer platform and Google AI Studio, in the coming weeks.
• Alphabet reports third-quarter results after the US market close on Oct 28.