Realtime AI News
Google DeepMind's Genie 4 world model adds real-time audio and on-the-fly steering
Google DeepMind released Genie 4, a world model that generates more photorealistic, interactive scenes with matching real-time audio and lets users rewrite the simulation with text prompts mid-run. It is available as a research preview for trusted testers, with a wider rollout to Google AI Ultra subscribers and Gemini API partners expected in the coming weeks.

Google DeepMind has released Genie 4, its newest world model, as a research preview for trusted testers. According to TechCrunch, the model generates the most photorealistic and interactive worlds the team has produced, and lets users type prompts mid-simulation to change what is happening on screen.
Audio is the most visible addition. Genie 4 generates real-time sound that matches the simulated world, including music, ambient noise and interactive dialogue. The model also loosens the editing controls: weather and camera angles can be changed live, and new people or objects can be added on the fly rather than only through preset actions.
Scene persistence has doubled. Genie 4 can hold a consistent scene for up to two minutes, twice the limit of Genie 3, which Google released as a research preview a little over a year ago. Physics are better and characters move in a more human-like way, closing two gaps that made earlier world models feel like demos.
The more consequential change is how Gemini fits in. Google's model acts as a director that plans steps and checks frames while Genie produces the visuals. In a research experiment tied to Gemini Robotics, a Gemini 4 Argon model broke a task into steps that Genie 4 simulated in real time, a possible route to testing robot scenarios in simulation before they happen in the physical world.
Tejas Kulkarni, a research director at Google, described the direction as "hybrid world models," with Gemini writing code and text while Genie ingests it and simulates reality at different levels of abstraction. That division of labor turns the world model from a video generator into something closer to a simulation kernel that can be instructed and orchestrated.
The pitched commercial uses remain gaming, interactive learning and robotics. One speculative example is stepping into a painting of a historical scene and talking to its subjects; a nearer-term one is film prototyping. Testers say the real-time loop matters most in ideation and storyboarding, where keeping a creative flow is the point. Engineer Diego Rivas said users may eventually use the multimodal Gemini Omni to turn those prototypes into production assets, though those capabilities are still a ways off.
The limits are explicit. Google's team acknowledges that details and objects drift the longer a simulation runs, and that it has not fully solved accurate audio or legible in-world text. Rivas stressed that it is still early days and that the end-to-end product experience is not yet defined.
Competition is crowding in. Runway shipped its second world model, GWM Worlds 2, a couple of months ago; World Labs is being acquired by AMD, Amazon is in talks to buy Decart, and Rivian just bought Third Dimension to bring simulation in-house for self-driving testing. Genie 4 is open now to trusted testers and will reach Google AI Ultra subscribers in Project Genie and early access partners on the Gemini API and AI Studio in the coming weeks.
Why it matters
World models are moving from showcase clips toward interactive simulation infrastructure, and putting Gemini in the director's chair makes it the orchestration layer for that simulation, and eventually for robot training. If two-minute consistency and audio accuracy keep improving, gaming, film previsualization and robotics testing are the first three workflows to change.
Nearby Updates
All10/28, 00:58
You don't have to use Meta's Muse for it to read your messages
Meta says its AI assistant Muse has passed 10 million users, about a quarter of whom use it daily. The company has also detailed how Muse handles data about people who never signed up, and the answer is that their messages, calendars and travel plans can be processed once a friend connects those accounts.
10/28, 01:07
Chargebacks911 Launches Indemnification for AI Agent Sales
Chargebacks911 has launched an indemnification offering for AI agent sales, per The Paypers, pushing the dispute specialist into agent-driven commerce. The product aims to turn unclear liability for agent-initiated transactions into a priced, insurable cost for merchants and platforms.
10/28, 01:08
Frontier labs say their models refuted the Exponential Time Hypothesis; TechCrunch asks an MIT scientist what that means
TechCrunch reports that frontier AI labs claim their models refuted the Exponential Time Hypothesis, a central conjecture in computational complexity theory, and asks an MIT computer scientist to explain OpenAI's discovery. The coverage frames the claim as still being digested rather than verified.
10/28, 01:16
Liquid AI ships LFM3: six 2B-class models on one backbone
Liquid AI has released LFM3, a family that branches six 2B-class models for text, vision, audio, omni and decision-making from a single 1.7B base. The company says LFM3-1.7B beats its larger LFM2.5-2.6B predecessor with 35% fewer parameters, and that every model fine-tunes on one GPU and runs on-device.