Imagine a computer game where the environment doesn't have to be pre-programmed. The world is created right before your eyes, it reacts to your decisions, and you can change it with a simple text command. This is the direction Google is taking with its new artificial intelligence model, Genie 4. The technology, unveiled on October 27, 2026, can create realistic, interactive worlds in real-time, including sounds, movement, and reactions from virtual characters. Although it is currently in the research phase, it could fundamentally change not only computer games, but also film production or robot training.
Google, through its research division DeepMind, has introduced a new generation of artificial intelligence called Genie 4. Unlike conventional image or video generators, this system can create environments in which the user can move and which continuously react to their actions.
This is not just about watching a pre-made animation. Genie 4 generates images and sounds simultaneously as the virtual world evolves. The user can look around, control the camera, and change the appearance of the surroundings. For example, if the user enters a command to start raining, the weather changes, or a new person appears in the scene, the artificial intelligence attempts to immediately implement the requested change.
Google refers to this technology as a world model, meaning a model of the world. Its purpose is not only to create a visually impressive scene, but also to simulate its further development.
A world that emerges during gameplay
While classic computer games use environments, objects, and rules created by developers, Genie 4 attempts to generate the appearance of the virtual world continuously. This opens up the possibility of creating interactive experiences whose specific course does not have to be predetermined.
During the presentation of the new technology, Google showed, for example, realistic natural landscapes, moving characters, or simulations of events that gradually evolve according to the given instructions.
A significant advancement is also the sound. Genie 4 can create sounds of the environment, music, or dialogues along with the image, so the resulting scene does not just appear as a silent virtual space.
An important part of the system is other artificial intelligence models from the Gemini family. The Gemini Omni technology helps to create simulations that correspond to the appearance and behavior of the real world, while other models ensure the logical sequence of individual events.
Google also uses the Gemini 3.8 Flash model, which can continuously monitor what is happening, plan individual steps, and check their execution. Thanks to this, the system attempts to maintain a meaningful development of the simulation even in situations where the user changes the original instructions.
The future of games without pre-made worlds?
The gaming industry is one of the areas where such technology could find significant applications.
Developing extensive game worlds today requires the work of graphic designers, programmers, animators, and other specialists. Generative models could significantly speed up some parts of this process, for example, when creating prototypes of environments or testing new game concepts.
In the longer term, we can imagine games whose environments would change according to the player's decisions much more freely than today. Instead of choosing from a few pre-made options, the player could influence the appearance and development of the virtual world with their own instructions.
However, this does not mean that classic computer games are ending. Genie 4 is not yet able to replace complex gaming systems with a long-term stable environment, sophisticated rules, and a vast storyline. It is more of a demonstration of technology that could one day become part of their development.
Two minutes of a virtual world and a number of limitations
Despite the impressive demonstrations, Genie 4 still has significant limitations. According to the technology website The Verge, the current generation can create interactive simulations lasting approximately two minutes. This is twice the duration of the previous Genie 3 model, but it is still insufficient for полноценное (full) gaming.
Google also warns that errors may occur during longer operation. Virtual objects may not always maintain the same appearance and location, the environment may lose its spatial coherence, and some physical interactions still do not correspond to reality.
Another problem is the accuracy of the generated sound or the display of readable text. Difficulties may arise, especially if the user makes several complex changes during one simulation.
Therefore, Google emphasizes that Genie 4 is still a research technology, not a finished platform for creating computer games.
It can also be used for robot training
Although the public's attention is likely to be drawn primarily to virtual worlds that resemble computer games, Google also sees potential for Genie 4 in other areas.
One of these is robotics. Developers could use the generated environments to train robots and autonomous systems, without having to test each scenario in the real world.
Artificial intelligence could, for example, create various traffic situations, work environments, or complex scenarios in which the robot learns to react to unexpected events.
Similar simulations can also be useful in the development of AI agents that are designed to perform more complex tasks independently.
Another application is in the film industry. Creators could use the technology to quickly prepare scene designs, test camera movements, or change the appearance of the environment before filming.
When will it be possible to try Genie 4?
Google has so far made Genie 4 available only to a limited group of trusted testers. The company wants to first gather feedback and verify the safety of the new technology.
Subsequently, it plans to provide access to Google AI Ultra subscribers through the Project Genie platform. It is also planning a preliminary release to selected partners through the Gemini API and Google AI Studio.
The exact date for wider availability has not yet been announced.
Google also stated that the image and audio content created by the Genie 4 model will contain a digital watermark called SynthID. This is intended to help identify materials created by artificial intelligence.
Genie 4 represents another step in the development of generative artificial intelligence. While systems that can create an image or short video based on a few words have recently been attracting attention, research is now moving towards environments that can be actively influenced.
Whether this technology will actually change the way computer games are created remains to be seen. However, it already suggests that the future of artificial intelligence may not only lie in answering questions or creating content. Its ability to simulate entire environments and react to events that occur within them may become increasingly important.
Jaroslav Holý
