BUSINESSCompanies · · Source: Google DeepMind
Google DeepMind unveils Genie 4: AI-generated worlds you can explore, with sound
On October 27, 2026, Google DeepMind introduced Genie 4, a so-called world model. It generates interactive environments you can change with language in real time – and, according to Google, it's the first to generate its own audio.
What a world model is
A world model doesn't produce a single image or video. It creates an environment you can move through and change – a bit like a video game, but without a pre-built world. Genie 4 builds on the Gemini model family.
What's new
- Language control: change the weather, move the camera, add people or objects while the simulation runs.
- Physics: basic physical effects such as airflow and collisions.
- Characters: people move more naturally and react to their surroundings.
- Sound: ambient noise and dialogue are generated natively – Google calls it the first real-time world model to do this.
Who can use it
Genie 4 is currently a research preview for trusted testers. Next in line are Google AI Ultra subscribers in “Project Genie” and early access partners of the Gemini API and Google AI Studio. There's a waitlist. All generated video and audio carry an invisible SynthID watermark.
Limitations
Google lists its own weaknesses: details and objects can drift over longer sessions, and complex physics, legible text and spatial accuracy are still research problems. Google says worlds stay coherent over “short sessions”.
What this means for you
- Marketing and product teams get a new way to make ideas visible: scenes, spaces or customer journeys can be played through before anything is built or filmed.
- It's not an everyday tool yet – access is limited to testers and, later, Google's most expensive subscription.
- The watermark shows that big vendors are increasingly labelling AI content by default.
Luigi's take
From a marketing perspective, this is a look at the future of the concept phase. Today it takes time and money to visualise an idea well enough for leadership to grasp it. If scenarios like a store concept, a trade-show booth or a customer experience can be played through with language, decisions get faster – and bad plans get cheaper.
Still, for most businesses this is something to watch, not to invest in. The limits Google itself names are significant. I'd rather put budget into AI tools that save time every day right now, and look at world models again in a year.
Source
Google DeepMind: Genie 4: a new generation of world simulation
This article was researched and written with AI assistance. Responsible: Luigi Lugmayr.
