Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering
- ID
- 40029
- Status
- summarized
- Published
- 28 Oct 2026, 12:30 AM
- Fetched
- 28 Oct 2026, 4:02 AM
- Provider
- TechCrunch
- Category
- technology
- Original URL
- https://techcrunch.com/2026/10/27/googles-genie-4-world-model-adds-audio-more-realistic-worlds-and-real-time-steering/
- Source URL
- https://techcrunch.com/feed/
Summary
- Score
- 5.5
- Created
- 28 Oct 2026, 4:03 AM
- Tags
- Audience
- developersai_ml_learnersai_agent_users
What happened
Google DeepMind released Genie 4 in research preview on Tuesday, adding real-time audio, mid-simulation text prompts, freeform edits like changing weather or camera angle, and up to two minutes of scene consistency — double Genie 3's limit. Gemini is integrated as a director that plans steps and checks frames, and a Gemini Robotics experiment had a Gemini 4 Argon model break tasks into steps that Genie 4 simulated in real time. The article also notes consolidation: Runway shipped GWM Worlds 2 months earlier, AMD is acquiring World Labs, Amazon is in talks to buy Decart, and Rivian acquired Third Dimension.
Why it matters
If you work on robotics/agent simulation or interactive prototypes, the jump from Genie 3's one-minute limit to two minutes plus real-time steering is the concrete detail to evaluate — but this is a research preview with no API, pricing, or general-availability details, so don't make production or budget decisions around it yet. For Malaysian builders, the text gives no local access, pricing, or infrastructure angle.
Discussion angle
Where is the line between a demo and a tool you can build on — does two-minute consistency plus audio make Genie 4 usable for agent/robot test loops, or is it still just a research preview until access, pricing, and latency are published?