Overview
- DeepMind announced Genie 4 on Tuesday, October 27, 2026, as a research preview of a “world model” that generates photorealistic, interactive 3D scenes in real time rather than pre‑rendered videos.
- The key new features are native audio generation that produces ambient sound, music and interactive dialogue together with visuals, roughly two minutes of scene coherence—double Genie 3—and clearer physics and more natural human motion.
- Genie 4 runs under tighter orchestration from Gemini variants: Gemini 3.8 Flash plans and checks frames to keep scenes coherent, Gemini Omni anchors appearance and behavior, and Gemini 4 Argon was shown breaking tasks into steps for robot‑simulation experiments.
- DeepMind says the release includes SynthID watermarking and safety guardrails, and it acknowledges persistent limits such as scene drift, imperfect audio and in‑world text accuracy, and short runtimes that keep Genie 4 in prototyping rather than production use.
- Google will expand access through Project Genie for Google AI Ultra subscribers and early Gemini API/AI Studio partners, a move that could speed prototyping for film, games and robot training and will be watched by investors after earlier Genie rollouts moved game and engine stocks.