Google DeepMind Releases Genie 4 World Model in Research Preview With Audio, Real-Time Edits

4.2 · 4d ago · US · AI

The Facts

  • Google DeepMind released Genie 4, its newest world model, as a research preview on Tuesday. ✓2 · TechCrunch, Ars Technica
  • Genie 4 is the first DeepMind world model with native audio, including ambient sound and interactive dialogue; TechCrunch also lists music. ✓2 · TechCrunch, Ars Technica
  • Users can change a simulation midstream, such as altering weather or camera angle or adding objects or people. ✓2 · TechCrunch, Ars Technica
  • DeepMind says Genie 4 has improved physics (e.g., aerodynamics and collisions) and more natural-looking human characters. ✓2 · TechCrunch, Ars Technica
  • Gemini is integrated with Genie 4 as a guiding agent that plans steps and checks frames. ✓2 · TechCrunch, Ars Technica
Dive deeper · 7 more factsShow less
  • Genie 4 can keep a scene consistent for up to about two minutes, double Genie 3's limit. ✓2 · TechCrunch, Ars Technica
  • Genie 4 is available now to trusted testers and will roll out in coming weeks to Google AI Ultra subscribers via Project Genie and to early access partners of the Gemini API and Google AI Studio. ✓2 · TechCrunch, Ars Technica
  • Google's stated use cases include gaming, interactive learning, robotics training and creative prototyping such as storyboarding. ✓2 · TechCrunch, Ars Technica
  • Google AI Ultra plans start at $100 per month; Gemini API pricing for Genie 4 is not yet available. Ars Technica only
  • Google acknowledged gaps: details and objects can drift over longer simulations, and audio accuracy and legibility of in-world text are unresolved. TechCrunch only
  • In a Gemini Robotics research experiment, a Gemini model broke tasks into steps that Genie 4 simulated in real time. TechCrunch only
  • Genie 3 was released as a research preview a little over a year ago. TechCrunch only

Context

Google DeepMind released Genie 4, a world model that generates interactive simulated environments frame by frame in real time, as a research preview. It adds native audio including speech, improved physics and human characters, and Gemini-guided steering of the simulation.

Access is initially limited to trusted testers, with a phased rollout to AI Ultra subscribers through Project Genie and to developers via the Gemini API and AI Studio. DeepMind says the model is still early-stage research.

Conflicting Reports

  • Gemini model naming differs: TechCrunch cites a 'Gemini 4 Argon' model in the robotics experiment; Ars Technica says Genie 4 uses Gemini 3.8 Flash as its guiding agent. These may refer to different uses.
  • Ars Technica says there is no specific timeline for wider access; TechCrunch says rollout to AI Ultra subscribers and API partners is expected 'in the coming weeks'.

How the Agent Panel Scored This

Independent AI personas each scored importance 1–10; the story's score is a trimmed mean (highest and lowest votes dropped). Vote spread: 2.

5 News Editor This is a concrete product launch from a major AI lab with notable capability advances, but it is a limited research preview with restricted access and no broad immediate impact on most people, so it sits in the moderate newsworthy range.
5 Science & Tech Analyst A notable research-preview release of a new world model with meaningful capability gains (audio, two-minute consistency, Gemini-guided steering), but it is limited to trusted testers with no broad immediate impact and is a product/research update rather than a peer-reviewed breakthrough.
4 Global Affairs Analyst A research-preview AI product release from a major company with limited, phased access and acknowledged gaps is a notable tech development but has limited broad geopolitical or societal impact.
4 Policy Expert A research-preview product launch by a major AI lab with limited access to trusted testers is a notable tech product release but has modest immediate impact on most people and no governmental or regulatory action.
4 Critical Reader A research-preview product release limited to trusted testers with acknowledged unresolved gaps is a concrete but niche AI product launch with limited immediate impact on most people.
4 Business Analyst A research-preview AI product launch with limited initial access and no pricing for developers is a notable tech release but has limited immediate market or economic impact.
3 Informed Citizen This is a product research-preview release with limited access to trusted testers and a later rollout to paid subscribers, so it has modest impact on ordinary people within a week.

Sources (2)

← All stories