Brief
Google launches Genie 4 AI model that creates realistic video and voice
The new Genie 4 model adds audio generation and improved human simulation to Google's AI world‑model lineup.
By Felo News Desk · Published
Google DeepMind introduced Genie 4, the latest version of its AI world model, as a limited research preview for a small group of early testers, Gizmodo reported on 27 October 2026.
Genie 4 builds on the earlier Genie 3 model unveiled in August 2025, adding the ability to generate human‑like audio that automatically matches any visual scene created by the user. The system can also produce photorealistic video that users can edit in real time with natural‑language prompts, and it simulates human motion, facial expressions and voices more fluidly than its predecessor.
Google’s blog post, cited by the article, says the model "unlocks unprecedented realism" and can simulate ambient sounds such as wind or waves, allowing AI‑generated characters to converse and react to events in a video. The company emphasizes that the model learns by observing the physical world, aiming to simulate real‑world physics, lighting changes and camera‑angle shifts.
According to the same blog post, Google views world models as a key path toward artificial general intelligence, echoing remarks from AI researcher Yann LeCun that world models may be more promising than large language models for AGI development.
Google says Genie 4 is being released with "strict safety guardrails" and has undergone "rigorous safety evaluations" to limit harmful content. The company acknowledges that the model can still struggle with visual consistency and audio accuracy in longer, more complex simulations. All generated media will be watermarked with a "SynthI" tag.
Key facts
- Google DeepMind launched Genie 4 as a limited research preview. (gizmodo.com)
- Genie 4 can generate photorealistic video and human‑like audio that matches the visual scene. (gizmodo.com)
- The model includes safety guardrails and will tag all output with SynthI. (gizmodo.com)
- Google says Genie 4 improves human motion, facial expression and voice realism over Genie 3. (gizmodo.com)
- Google cites world models as a route toward artificial general intelligence. (gizmodo.com)
Sources
- [1] gizmodo.com — originally reported as “Google's New AI World Model Can Generate Human Voices, Not Just Video”
