Updated on October 28, 2026
Genie 4 is Google DeepMind's real-time world model, announced on October 27, 2026, and the first in the Genie line to produce native audio alongside interactive environments. You steer scenes with keyboard controls and natural-language prompts while ambient sound, music and character dialogue generate in the same pass as the visuals. Access is limited to a trusted-tester research preview; broader availability is planned through Project Genie for Google AI Ultra subscribers at $200/month and via Gemini API partners.
Genie 4 renders each frame on the fly in response to keyboard inputs and text prompts. Gemini Omni grounds the scene's look in real-world appearance, while Gemini 3.8 Flash acts as a guiding agent: it breaks goals into steps, checks every output frame for consistency and rewrites the plan when conditions change. Between your prompts, Gemini keeps the world moving on its own.
That loop is what separates a world model from a standard video generator like Veo 3. Type 'add a thunderstorm' mid-session and the sky darkens, rain sounds appear, lighting shifts. The environment reacts rather than replaying a clip. DeepMind has also paired Gemini 4 Argon with Genie 4 to simulate Apptronik's Apollo humanoid playing checkers, feeding each sub-task as a text caption while the scene renders, a step toward training robots inside computer simulation.
DeepMind's trusted-tester pool spans filmmaking, game development, journalism and engineering. Filmmakers treat Genie 4 as a draft mode for pre-visualization: step into a live scene, run multiple takes, adjust lighting and dialogue on the fly. Game designers jump from a text description to an explorable prototype in seconds, skipping engine code entirely.
Photojournalists have turned flat archival photos into walkable reconstructions of historical events. Enterprise partners are evaluating the model for rapid scenario generation in manufacturing and logistics.
| Generation | Date | Key advance |
|---|---|---|
| Genie 1 | March 2024 | 2D worlds from a single image |
| Genie 2 | December 2024 | 3D environments, 10-20 s sessions |
| Genie 3 | August 2025 | Real-time at 720p / 24 fps, ~1 min coherence |
| Genie 4 | October 2026 | Native audio, live text steering, doubled coherence |
Genie 4 has no standalone price. Access is planned through two channels: Project Genie for Google AI Ultra subscribers, and the Gemini API / AI Studio for early-access partners. A waitlist is open. Project Genie currently runs on Genie 3 and is available in 140+ countries for adults with an AI Ultra plan. No country list specific to Genie 4 has been published yet.
The $200/month AI Ultra tier is the only plan that includes Project Genie. A lower $100/month Ultra tier exists but without Project Genie access. Google AI Pro starts at $19.99/month. Pricing may change; Google's official plans page is the definitive source.
The world-model field is moving fast. Runway shipped GWM Worlds 2, a playable world model with audio, a few months prior. AMD signed a definitive agreement to acquire World Labs (Marble) for roughly $8.2 billion in September 2026, pending regulatory approval. ByteDance is reportedly building a Seedance-based world model to compete directly with Genie.
Genie 4 is Google DeepMind's real-time world model, announced October 27, 2026. It generates interactive environments with native audio, including ambient sound, music and character dialogue. Built on the Gemini family, it uses keyboard controls and text prompts for navigation and scene editing. All outputs carry a SynthID watermark. It launched as a research preview for trusted testers.
Genie 4 has no standalone price. Access is planned through Project Genie, included only in the $200/month Google AI Ultra tier. A $100/month Ultra tier exists but excludes Project Genie. Early-access partners of the Gemini API and AI Studio will also get access. No free tier or trial has been announced for Project Genie.
Today you can use Genie 3 inside Project Genie at labs.google as an 18+ Google AI Ultra subscriber. Genie 4 is in research preview with a waitlist on the DeepMind blog. It will arrive in Project Genie and through Gemini API partners in the coming weeks, though no exact date has been confirmed.
Project Genie is Google Labs' research prototype that lets AI Ultra subscribers create and explore 60-second interactive worlds from text, images or Street View. It launched January 29, 2026 in the United States and expanded globally on May 19, 2026, covering 140+ countries. It currently runs on Genie 3, with Genie 4 announced as the next model update.
Verdict: For game studios prototyping a concept before writing engine code, filmmakers blocking out scenes without a set, or robotics teams that want safe training environments, Genie 4 opens a rare playground. The AI Ultra price tag and the waitlist are real barriers, but the jump from silent worlds to sound-generating, promptable scenes marks a turning point for interactive world models.
« An all-in-one creative platform that brings together more than 30 image and video models in a single interface. It includes camera controls that let you direct like a real filmmaker, characters that remain consistent from scene to scene, and ready-to-use tools for ads and short-form content »
Project Genie 
Oasis AI Game 
Genie 3 by Google 
