Google DeepMind announces Genie 4, a new world model for real-time multimodal interactive AI experiences with audio and steering
Model release ● Confirmed 90% confidence first seen
Google DeepMind unveiled Genie 4, an updated world model designed to generate real-time interactive experiences with improved realism, including more realistic physics and humans, and added audio generation. The model supports longer simulations (up to about two minutes), allows users to steer experiences during simulation with natural-language prompts, and is rolling out first as a research preview for trusted testers before broader access.
DeepMind
Genie 4
Google
Google DeepMind
AI Studio
Gemini 3.8 Flash
Gemini API
Genie 3
Google AI Studio
Google AI Ultra
Project Genie
SynthID
Decision brief
- What changed
- Google DeepMind announced Genie 4, a natively multimodal world model that generates real-time interactive simulations with photoreal visuals, physics-consistent movement, and audio. Compared with Genie 3, it extends coherent experience length to about two minutes and adds natural-language steering during an ongoing simulation; rollout starts as a research preview for trusted testers before expanding to Google AI Ultra subscribers, Gemini API early access, and Google AI Studio.
- Why it matters
- For leaders evaluating AI product roadmaps, Genie 4 signals that Google is moving world-model capabilities closer to usable interactive experiences rather than static media generation. The combination of longer coherent sessions, live prompt steering, and planned API access could create new options for prototyping simulation, entertainment, training, and immersive customer experiences, but access is initially limited and staged.
- Evidence
- All three cited sources report the same core facts: Genie 4 was announced by Google DeepMind, supports roughly two-minute interactive simulations, adds real-time prompt steering, and includes audio generation. The reporting is consistent across Google's own announcement and independent coverage from The Verge and TechCrunch, though most technical detail and rollout specifics originate from Google.
- What remains uncertain
- The coverage does not establish production reliability, pricing, latency, developer constraints, safety guardrails beyond SynthID watermarking, or how broadly available the API access will be at launch. It is also not yet verified from independent testing how well Genie 4 performs across real business use cases versus curated demos and trusted-tester feedback.
- Monitor next
- Watch for Gemini API and Google AI Studio early-access documentation or trusted-tester results that clarify latency, pricing, capability limits, and supported enterprise use cases.
Analytical support, not advice — assumptions and open questions stated above.
Source coverage
Google’s next Genie AI model generates more realistic experiences
The Verge
independent
DeepMind unveils Genie 4 world model promising more realistic physics and humans
Ars Technica
independent
Genie 4: A new generation of world simulation
Google
official
Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering
TechCrunch
independent