Genie (Google DeepMind)
Google DeepMind's real time world model, currently a waitlist only research preview
A correction first: the URL originally given for this tool, deepmind.google, is DeepMind's general homepage, not a dedicated page for Genie. The actual model page lives at deepmind.google/models/genie, which is the URL used for this listing. Genie, now on its fourth version, is Google DeepMind's world model, described by the company as its most realistic and interactive world model yet. It generates explorable worlds from text and image prompts, then lets you change that world in real time using natural language, moving the camera, adding characters, or changing the weather as the simulation keeps running. It simulates physical behavior like objects colliding, water flowing, and trees swaying, generates characters that move and behave naturally, and can produce native audio alongside the visuals. Cause and effect event sequences are supported too, with Gemini models helping keep everything coherent over time. DeepMind is exploring it for robotics simulation and for training agents such as SIMA. This needs to be said plainly: Genie 4 is a research preview, not a publicly available product. Access is currently limited to trusted testers through a waitlist linked from the page. Broader access is planned eventually for Google AI Ultra subscribers through something called Project Genie, plus early access partners of the Gemini API and Google AI Studio, but no start date has been announced and no pricing exists yet since there is no general public release.
- Category: AI Simulation
- Platforms: Web
Key features
- Generates explorable worlds from text and image prompts
- Real time natural language world editing, including camera movement, adding characters, and changing weather
- Simulated physical behavior, including collisions, water flow, and swaying trees
- Characters that move and behave naturally, with native audio generation
- Cause and effect event sequencing, with Gemini models helping maintain coherence
- Being explored for robotics simulation and for training agents such as SIMA
Pros & Cons
Pros
- Notably ambitious technical scope, combining real time world generation, physics simulation, native audio, and natural language editing in one system
- Directly tied to Google DeepMind's broader Gemini and SIMA research rather than being a standalone side project
- Being explored for genuinely practical downstream uses like robotics simulation and agent training, not just demos
Cons
- Not generally available: Genie 4 is explicitly a research preview, with access currently limited to trusted testers through a waitlist
- No pricing has been announced since there is no general public access yet
- Broader access is only planned, for Google AI Ultra subscribers through Project Genie and select Gemini API and Google AI Studio partners, with no announced start date
- The originally requested URL pointed to DeepMind's general homepage rather than Genie's own dedicated model page
FAQs
Can I use Genie right now?
Not generally. It is currently a research preview limited to trusted testers accessed through a waitlist on DeepMind's site.
When will Genie be publicly available?
No public release date has been announced. Broader access is planned eventually for Google AI Ultra subscribers and select API partners.
Does Genie cost anything?
No pricing has been published since there is no general public access to the product yet.