Genie 4

What it does

Genie 4 is Google DeepMind's interactive world model. A text or image prompt starts a scene you can move through, and a later prompt can change the weather, the camera, or who is in it while it runs. The 27 October 2026 post says the model generates native audio, including ambience and spoken dialogue, and that Gemini models help keep the simulation coherent. It is a research preview for trusted testers. DeepMind says a wider release starts with Google AI Ultra subscribers in Project Genie and early-access partners on the Gemini API and Google AI Studio.

Where it fits

Read the model page, then join the waitlist if you want access. With access, start from a still or a short prompt, steer with the keyboard and text, and stop when details drift. Treat the result as a simulation, not a finished shot.

Before you commit

This is a research preview, not a public app. DeepMind says detail and object permanence still drift on longer sessions, after several edits, or as you move through the scene, and that text, physics, and audio accuracy are unfinished. Picture and sound include a SynthID watermark. This check did not run the model.

Release & update notes

Dates below refer to the linked announcement or Product Hunt launch. They are separate from the date this profile was added.

• Official announcement / update · 2026-10-27 ↗

  Research preview. Interactive world simulation with native audio.