What Genie 4 can do
Google DeepMind has presented Genie 4 as a research preview. According to heise online, the world model generates interactive 3D environments from text or image prompts that users can explore with a keyboard. The new feature is audio generation: alongside images, Genie 4 produces matching ambient sounds, music and speech. According to the report, users can also change the weather or add objects while the simulation is running.
Limits and access
According to Bloomberg, simulations are currently limited to up to two minutes, and the environments are not consistently coherent throughout, as Google researcher Hang Qi acknowledges. Golem reports that the model is initially available to selected testers. Heise names the next steps as access via Project Genie for Google AI Ultra subscribers and selected developers via the Gemini API and Google AI Studio, without giving a date for general availability.
Context
What Genie 4 means in practice is still open, according to DeepMind product manager Diego Rivas, as the technology remains in the research phase. The reported use cases, such as film scenes, game prototypes and training robots and AI agents, are so far trials rather than established applications. For decision-makers, the key point to watch is whether world models will eventually help plan actions virtually in advance.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
