Google DeepMind unveiled its latest world model, Genie 4, in a research preview on Tuesday. This new model is capable of generating the most photorealistic and interactive environments to date, enabling users to input prompts during simulations to modify the scene. Additionally, the model produces real-time audio that corresponds with the simulated environment, encompassing music, ambient sounds, and interactive dialogue.
The introduction of Genie 4 follows the recent launch of AI video startup Runway’s second world model, GWM Worlds 2, which is a playable model that dynamically responds to user interactions and includes audio generation. Concurrently, the world model sector is experiencing a wave of consolidation, with Fei-Fei Li’s World Labs being acquired by AMD, Amazon negotiating to purchase Decart, and Rivian acquiring Third Dimension to enhance its in-house simulation capabilities for self-driving tests.
Genie 4 represents Google’s strategic effort to maintain its competitive edge in a rapidly evolving landscape. The company released its predecessor, Genie 3, as a research preview just over a year ago. In addition to its enhanced photorealism and audio features, Genie 4 permits freeform modifications, such as altering weather conditions or camera angles in real time, as well as adding new characters or objects. It can sustain a consistent scene for up to two minutes, which is double the duration of its predecessor, and offers improved physics and more lifelike characters.
Google’s AI model, Gemini, is now more closely integrated with Genie 4, functioning as the director that orchestrates steps and verifies frames while Genie generates the visuals. In a research experiment associated with Gemini Robotics, a Gemini 4 Argon model decomposed tasks into steps that Genie 4 simulated in real time. This experiment illustrates a potential method for testing robotic scenarios in a simulated environment prior to real-world implementation.
Tejas Kulkarni, research director at Google, commented during a briefing, “What we’ll increasingly see in the future are these hybrid world models, where you have a combination of Gemini writing code and text, and Genie ingesting that in real time and simulating reality at different levels of abstraction.” As with other world models, the proposed commercial applications include gaming, interactive education, and robotics. One imaginative example of interactive learning could involve entering a painting of a historical event and conversing with the depicted figures about the occurrence.
However, “trusted testers” — individuals who gain early access to Google’s innovative products — are suggesting a potential new application: film prototyping. Diego Rivas, a Google DeepMind engineer, stated during a press briefing, “The real-time aspect becomes most valuable in the areas where maintaining a state of flow in the creativity space becomes incredibly useful. For instance, during the ideation phase or the storyboarding phase — anytime you’re in need of rapid ideation.”
Google also provides Veo, its AI video generation model that creates high-definition clips with synchronized audio from text, image, or video prompts — a tool that has gained popularity among filmmakers. Rivas further mentioned that users could eventually depend on Gemini Omni, a multimodal large language model capable of generating video from images, audio, and text, to transform those prototypes into creative assets, although he noted that such capabilities are still some time away.
Throughout the briefing, Rivas emphasized that Genie 4 is still in its early stages, and the complete end-to-end product experience remains uncertain. The team also recognized that Genie 4 has limitations: details and objects may drift as a simulation progresses, and they have yet to fully resolve issues regarding audio accuracy and the legibility of in-world text.
Genie 4 is currently accessible as a research preview for trusted testers. In the upcoming weeks, the model will be made available to Google AI Ultra subscribers participating in Project Genie, Google’s AI tool for generating interactive worlds from text or image prompts, as well as to early access partners of the Gemini API and Google AI Studio.
