Genie 4: Audio, Two Minutes, and a Typo Worth Reading
Google's blog post announcing Genie 4 is 246 words. The second sentence is broken. "Genie 4 is natively multimodal, enabling audio-like music, ambient noises, and interactive dialogue to be simultaneously generated with the world." Audio-like music. Music that is similar to audio. Read it as an em dash pair that somebody's de-AI pass flattened into a hyphen, and the original was probably "audio,…
Google announced Genie 4 on 27 October 2026. Genie 4 is multimodal, capable of generating audio, music, ambient sounds, and interactive dialogue simultaneously with the world. It features native audio generation, enabling sound design to be built directly into a simulation without the need for separate audio creation. The improvement over Genie 3 is about two minutes of coherent output, compared to one minute previously.
Genie 4 offers steering capabilities to change weather, camera, and cast during mid-simulation. Physics in Genie 4 includes better collisions, aerodynamics, and human motion compared to its predecessor. Access is limited to trusted testers now, with Ultra subscribers expected soon. Pricing has not been disclosed. Google specifies "short sessions" rather than the two-minute benchmark reported by other sources.
The audio generated by Genie 4 is considered a significant advancement, as it eliminates the need for a separate sound designer to build environmental sounds like rain, footsteps, and room tones. However, the demos of Genie 4's audio and dialogue generation are not entirely accurate, with some users receiving unexpected or incorrect outputs.
The overall coherence and fidelity of the generated audio and visuals are described as promising, but there are still issues with detail drift and accuracy when camera movements or prompt edits are made. The integration of a second model into the Genie 4 loop, Gemini Omni, anchors the simulation in realistic visuals and motion, while Gemini 3.8 Flash acts as a planner, refining the simulation's performance based on feedback.
An example of Genie 4's capabilities is demonstrated in a checkers game played by an Apptronik Apollo humanoid, which showcases the potential for more advanced robotics applications. However, the checkers demo has issues, with physics inaccuracies observed in the demo. The Hacker News discussion surrounding Genie 4 included debates about Google's streaming service, Stadia, and its relevance to the technology's potential consumer product impact.
Overall, Genie 4 represents an important step forward in the development of world models and their applications, particularly in the field of robotics and simulation.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
