Google Launches Genie 4 for AI Ultra Subscribers With Real-Time Audio and World Editing
Summary: Google has released Genie 4 to eligible AI Ultra subscribers via Project Genie, enabling real-time text-based modification of interactive 3D worlds with native audio. The update extends Genie 3 sessions to three minutes and introduces a community gallery.
Google is rolling out Genie 4 to eligible AI Ultra subscribers through Project Genie, marking the first public access to its latest interactive world model since its initial unveiling in October Original reporting. The company announced the availability of the new model on Wednesday, November 18, 2026, expanding the browser-based experiment that allows users to generate and explore digital environments announced Wednesday.
Real-Time World Editing and Native Audio
Genie 4 represents a significant shift from previous iterations by allowing users to modify generated worlds during active exploration. Instead of rendering a static video from a prompt, the model generates scenes dynamically as users navigate them. Participants can type instructions midway through a session to alter environmental conditions, such as changing the weather, introducing new characters, or redirecting narrative events without restarting the simulation Original reporting.
For the first time, Genie 4 pairs these visual simulations with real-time native audio. The model generates ambient soundscapes, such as wind and footsteps, alongside interactive character dialogue that reacts to events within the scene announced Wednesday. Google states that the model improves the realism of human movement and object physics, though these claims are based on internal demonstrations rather than independent third-party testing Original reporting.
The technology relies on a combination of advanced world models and the Gemini family of AI systems. Gemini Omni anchors the simulations in real-world appearance and behavior, while other Gemini models provide the complex reasoning needed to maintain a simulation’s coherence and continuity unveiled in October. This integration allows the system to simulate how people move, react, and speak in response to user inputs, creating a more immersive experience than prior versions.
Subscription Access and Pricing
Access to Genie 4 is restricted to subscribers of the Google AI Ultra tier. In the United States, this subscription costs $199.99 per month costs $199.99. Google confirmed that eligible Ultra users can access the new experience immediately, while subscribers to the lower-tier AI Pro plan are currently limited to joining a waitlist Original reporting. The company has not provided a timeline for when Genie 4 will become available to Pro subscribers beyond the existing waitlist mechanism.
The Google AI Ultra plan offers the highest level of access to the company’s AI features, including up to 20 times higher usage limits than the Pro plan and priority access to new innovations costs $199.99. This tiered access strategy highlights Google’s approach to managing the computational demands of real-time world simulation, which requires significant processing power compared to standard text or image generation tasks.
Updates to Genie 3 and New Creation Tools
Alongside the launch of Genie 4, Google is maintaining support for Genie 3 within Project Genie, offering users a choice between the newer model’s enhanced visuals and audio or the older model’s longer session capabilities. Genie 3 sessions have been extended to three minutes, up from the previous 60-second limit, and now support seeding worlds from 360-degree panorama images announced Wednesday. Users can also link multiple Genie 3 worlds together to create longer continuous journeys.
The update introduces a new Project Genie agent, an AI creation assistant that helps users build visions through natural language conversation. This agent orchestrates Google’s multimodal models to establish visual styles and custom background scores. Additionally, a new community gallery allows users to discover, explore, and remix worlds created by others, including curated collaborations with creators such as technologist Will aka BLVCKL!GHT and filmmaker Anand Gandhi announced Wednesday.
Safety Measures and Technical Limitations
Google acknowledges that Genie 4 remains an experimental research prototype. The technology differs from traditional game engines, as it predicts visual and auditory outputs rather than relying on stable, rule-based levels. Consequently, objects may drift between views, complex physics interactions can be inconsistent, and generated speech may contain inaccuracies Original reporting.
To address safety concerns, Google stated that all video and audio generated within Project Genie includes SynthID watermarks. Furthermore, all model inputs and outputs undergo safety screening processes announced Wednesday. The company continues to cite challenges such as camera latency and long-term scene permanence as areas for ongoing research and development.
The evolution of the Genie model series reflects Google’s broader efforts in artificial general intelligence (AGI) research. Understanding and simulating the world is considered a crucial aspect of building AGI, as it enables the creation of safe, infinitely variable simulations for training robots and agents unveiled in October. From the initial 2D side-scrolling experiments in 2024 to the current 3D interactive environments, the project has progressively increased in complexity and realism Original reporting.