genie4.art is an independent guide to Genie 4, the natively multimodal world model Google DeepMind announced on October 27, 2026. Instead of rendering a fixed clip, Genie 4 builds a playable environment that you steer in real time with plain language, complete with physics, ambient audio, music and interactive characters.
What the site covers
- Features: real-time natural-language steering, persistent worlds, physics and human realism, SynthID watermarking on every generated frame.
- Genie 3 comparison: what actually changed between generations, and which limits still apply.
- Examples: the published demo worlds, with notes on what each one demonstrates.
- Access: how testing opened up, Project Genie for Google AI Ultra subscribers, and the Gemini API and AI Studio routes.
- Prompting guide: practical world-prompt patterns for keeping a scene coherent as you steer it.
Who it is for
Researchers, filmmakers and game developers evaluating world models, and anyone deciding whether Genie 4 fits a pre-visualisation, prototyping or simulation workflow.
Why it is different
Most coverage is a single announcement post. genie4.art keeps a structured, plain-language reference that is updated as DeepMind publishes new capabilities, and it is clearly labelled as unofficial third-party analysis rather than Google documentation.
