Wikipedia

Genie (world model)

Genie, Genie 2 and Genie 3 are world models developed by Google DeepMind that can generate game-like, interactive virtual worlds based on text, images, or sketches. Genie 3 is avialable in the form of Project Genie to Google AI Ultra[a] subscribers via Google Labs.

Genie
Original authorGoogle DeepMind
Initial releaseJanuary 29, 2026; 4 months ago (2026-01-29)
LicenseProprietary
Websitelabs.google/projectgenie[1]

The first version of Genie was introduced in March 2024, and was capable of generating two-dimensional interactive environments. Its successor, Genie 2, released in December of that year, expanded these capabilities to generate three-dimensional environments. Genie 3 was introduced in August 2025, with higher-resolution world generations and increased memory capability. In February 2026, Waymo adopted Genie 3 and created a specialized world model, the Waymo World Model, for autonomous driving simulation.

Genie has been used for training AI agents in three-dimensional environments and video game design. The developers of Genie are not concerned with video game development. It can generate three-dimensional environments in 720p, at 24 frames per second. It is also capable of generating streets via an integration with Google Street View.

History

edit

The first version of Genie was introduced in March 2024.[2] Its successor, Genie 2, was introduced in December, introducing the capability of generating three-dimensional environments.[3]

In August 2025, Genie 3 was introduced, introducing higher-resolution simulations and increased memory from Genie 2, which lost its memory after only ten seconds.[4][5]

On January 29, 2026, Genie 3 was released to the public in the form of Project Genie, which was only available to Google AI Ultra[a] subscribers in the United States who are over 18 years old.[6][7] After the release of Project Genie, shares of video game producers decreased significantly, with Take-Two Interactive decreasing by 9.3%, Roblox Corporation decreasing by 15%, Unity Software decreasing by 21%, and CD Projekt decreasing by 8%.[8]

In February 2026, Waymo adopted a fine-tuned version of Genie 3, the Waymo World Model, for simulating edge case scenarios to train their robotaxis.[9]

In May 2026, Genie gained an integration with Google Street View, increasing its capabilities to generate streets.[10] Project Genie was also made available to Google AI Ultra[a] subscribers over 18 worldwide.[11]

Models

edit

Original

edit

The original Genie model was introduced in March 2024, and was only capable of generating two-dimensional environments. It generates frames of the world from interactive input and previous frames, and was trained off of thousands of hours of video game footage.[2] The original Genie generated worlds at only one frame per second, which made worlds unplayable as video games.[3]

Genie 2

edit

Genie 2 was introduced in December 2024, and had the capability of generating three-dimensional environments as well as two-dimensional ones. According to Google, Genie 2 only generates a consistent and plausible world for up to a minute, but Ars Technica reported that it could only do so for up to 10–20 seconds.[3] Genie 2 could only generate worlds in 360p resolution.[12]

Genie 3

edit

Genie 3 was introduced in August 2025, and is capable of generating and modifying interactive worlds in real-time. It has a memory of one minute, which is higher than Genie 2's 10–20 seconds.[4] Genie 3 generates worlds in 720p at 24 frames per second.[4][13] Genie 3 has an integration with Google Street View, which allows it to generate streets.[14]

Waymo World Model

edit

The Waymo World Model is a modified version of Genie 3 created by Waymo in February 2026. It is used for simulating edge cases for training Waymo's self-driving cars. The edge case scenarios are not commonly encountered in real life, and include situations such as sudden tornadoes, elephant encounters, and snow on the Golden Gate Bridge.[9][15] Waymo also created a smaller version of the Waymo World Model that uses less computation.[16]

It was adopted in February 2026.[9] The Waymo World Model helped Waymo introduce its self-driving cars to 11 cities in the United States.[10]

The Waymo World Model is designed to be realistic, and outputs lidar at four times the speed of Genie 3.[9][15] It is capable of converting videos recorded with a dashcam or phone into worlds.[17] It has three types of simulation controls, being driving action control, scene layout control, and language control. Driving action control controls how it responds to inputs, scene layout control controls the generated scenario, and language control controls environmental conditions.[18][19]

Project Genie

edit

Project Genie is a website that allows Google AI Ultra[a] subscribers and are over the age of 18 to access Genie 3.[11] Project Genie was released on January 29, 2026.[6] Project Genie is accessible via Google Labs.[1]

Project Genie has been used for training AI agents in interactive environments and video game design.[21] It has been used to create imitations of Nintendo games, such as Super Mario 64 and The Legend of Zelda: Breath of the Wild.[22] According to Genie's developers, the main focus of Genie is in robotics and simulations, not in video games.[14]

It has a 60-second limit on world exploration.[1][6] In generated worlds, the WASD keys are used to move, the arrow keys are used to turn the camera, and the space bar is used to ascend.[23]

After the release of Project Genie, shares of video game producers decreased significantly.[8]

See also

edit

Notes

edit
  1. 1 2 3 4 Costs $249.99 per month.[20]

References

edit
  1. 1 2 3 Vigliarolo, Brandon (January 29, 2026). "Google's Project Genie could put even more game developers out of work". The Register. Archived from the original on February 18, 2026. Retrieved April 13, 2026.
  2. 1 2 Orland, Kyle (March 5, 2024). "Google's Genie game maker is what happens when AI watches 30K hrs of video games". Ars Technica. Archived from the original on December 8, 2024.
  3. 1 2 3 Orland, Kyle (December 6, 2024). "Google's Genie 2 "world model" reveal leaves more questions than answers". Ars Technica. Archived from the original on December 13, 2024.
  4. 1 2 3 Whitwam, Ryan (August 5, 2025). "DeepMind reveals Genie 3 "world model" that creates real-time interactive simulations". Ars Technica. Archived from the original on March 24, 2026. Retrieved April 15, 2026.
  5. ↑ Bellan, Rebecca (August 5, 2025). "DeepMind thinks its new Genie 3 world model presents a stepping stone toward AGI". TechCrunch. Archived from the original on August 5, 2025.
  6. 1 2 3 Whitwam, Ryan (January 29, 2026). "Google Project Genie lets you create interactive worlds from a photo or prompt". Ars Technica.
  7. ↑ Bonifacic, Igor (January 31, 2026). "Google's Project Genie lets you create your own 3D interactive worlds". Engadget. Retrieved April 12, 2026.
  8. 1 2 Pernell, Avalon (January 30, 2026). "Unity, Video Game Stocks Fall as Google's AI Tool Sparks Fears". Bloomberg News. Retrieved April 14, 2026.
  9. 1 2 3 4 Martindale, Jon (February 6, 2026). "Waymo Is Using Google's Genie 3 AI to Practice Handling Tornadoes, Elephants". PCMag. Retrieved April 6, 2026.
  10. 1 2 Bellan, Rebecca (May 19, 2026). "Google's Genie world model can now simulate real streets with Street View". TechCrunch. Archived from the original on May 19, 2026. Retrieved May 26, 2026.
  11. 1 2 Conditt, Jessica (May 19, 2026). "Project Genie adds Google Street View integration and goes live for global AI Ultra users". Engadget. Retrieved May 27, 2026.
  12. ↑ Kan, Michael (August 6, 2025). "Google's Genie 3 Hints at a Future Where AI Builds the Video Games We Play". PCMag. Retrieved April 12, 2026.
  13. ↑ Dredge, Stuart (August 6, 2025). "Google DeepMind's Genie 3 uses AI to make game-like worlds". Music Ally. Archived from the original on August 19, 2025. Retrieved April 21, 2026.
  14. 1 2 Kirchner, Malte (May 25, 2026). "Like 2021 with LLMs: Google researchers on the future of world models". heise online. Retrieved May 27, 2026.
  15. 1 2 Whitwam, Ryan (February 7, 2026). "Waymo leverages Genie 3 to create a world model for self-driving cars". Ars Technica. Retrieved April 14, 2026.
  16. ↑ Bastian, Matthias (February 6, 2026). "Waymo taps Google Deepmind's Genie 3 to simulate driving scenarios its cars have never seen". The Decoder. Archived from the original on February 7, 2026. Retrieved April 15, 2026.
  17. ↑ Burnett, Steward (February 9, 2026). "Waymo unveils DeepMind-powered world simulation model". Automotive World. Archived from the original on February 10, 2026. Retrieved April 15, 2026.
  18. ↑ Sprigg, Sam (February 12, 2026). "Waymo Launches Generative 3D World Model to Train Self-Driving Cars". Auganix. Archived from the original on April 15, 2026. Retrieved April 15, 2026.
  19. ↑ Hawkins, Andrew (February 6, 2026). "What happens when Waymo runs into a tornado? Or an elephant?". The Verge. Retrieved April 15, 2026.
  20. ↑ Dredge, Stuart (February 2, 2026). "Google's Project Genie creates AI-generated interactive worlds". Music Ally. Archived from the original on March 9, 2026. Retrieved April 21, 2026.
  21. ↑ Martindale, Jon (January 30, 2026). "Google's Project Genie Tool Lets You Build Virtual Worlds for Training or Just Fun". Archived from the original on April 11, 2026. Retrieved April 6, 2026.
  22. ↑ Peters, Jay (January 29, 2026). "Google's AI helped me make bad Nintendo knockoffs". The Verge. Retrieved April 6, 2026.
  23. ↑ Bellan, Rebecca (January 29, 2026). "I built marshmallow castles in Google's new AI-world generator". TechCrunch. Retrieved April 13, 2026.
edit
Retrieved from "https://en.wikipedia.org/w/index.php?title=Genie_(world_model)&oldid=1356869858"