Google DeepMind has unveiled Genie 4, a groundbreaking new world model available in research preview. This advanced model creates exceptionally photorealistic and interactive simulated environments. Users can now modify scenes mid-simulation by simply typing prompts. Genie 4 also produces real-time, contextually appropriate audio, including music and dialogue. It significantly surpasses its predecessor, Genie 3, in photorealism and interactive capabilities. The new model allows for freeform edits, such as altering weather or camera angles dynamically. Users can also introduce new characters or objects into the simulation in real time. Genie 4 can sustain scene consistency for up to two minutes, a notable improvement over the previous version. It boasts enhanced physics and more lifelike characters. Google anticipates a hybrid future where Gemini and Genie models collaborate to simulate reality.