# Genie 4, explained

> Google DeepMind's world model (announced October 27, 2026) that turns a text prompt or an image into an interactive scene you can walk through, hear, and change by typing while it runs. Research preview for trusted testers only; next are Google AI Ultra subscribers in Project Genie and early access partners of the Gemini API and Google AI Studio, with no date or price published. Independent explainer, not affiliated with Google. Updated 2026-10-28. Full page: https://lab.awaara.cn/genie-4/

## What Genie 4 is

- A world model, not a video generator. A video generator renders a clip you watch. Genie 4 renders a scene frame by frame as you move through it with keyboard controls, and it keeps responding to what you do and type.
- Natively multimodal: music, ambient noise, and interactive dialogue are generated at the same time as the scene, not dubbed over it. Characters converse and react to inputs or events in the world, in real time.
- Live steering: type a change mid simulation (new weather, a different camera angle, a new person or object) and Gemini rewrites your request to fit the scene, then keeps the scene progressing between inputs.
- Gemini in the loop: Gemini Omni anchors the scene in real world appearance and behavior. Gemini 3.8 Flash acts as a guiding agent, watching frames, breaking goals into steps, and checking each step worked. DeepMind has also paired Genie 4 with Gemini 4 Argon for robotics work, for example rendering Apptronik's Apollo humanoid playing checkers from Argon's step breakdown.

## Genie 4 vs Genie 3

- Preview: Genie 3 in August 2025; Genie 4 on October 27, 2026.
- Scene length: about 1 minute for Genie 3 (reported), about 2 minutes for Genie 4 (reported by TechCrunch; DeepMind's own post says only "short sessions").
- Audio: Genie 3 scenes were silent; Genie 4 generates native audio, music, and dialogue.
- Live edits: no mid scene edits at Project Genie launch for Genie 3; Genie 4 allows typed edits mid scene.
- People: Genie 4 characters move naturally, react to surroundings, and can be steered live, which DeepMind calls the most dramatic leap.

## Access checker

- Trusted testers: hold the research preview now, through invitation, not through Project Genie.
- Google AI Ultra subscribers (US): next in line, via the Project Genie waitlist. No date is published.
- Google AI Ultra subscribers outside the US: not announced yet. Project Genie is US only for Ultra subscribers so far, and no Genie 4 country list is published.
- Gemini API and Google AI Studio early access partners: in the partner wave, by invitation, with no API price published.
- Everyone else: no public access yet. The only official step is the Project Genie waitlist.
- Waitlist: https://forms.gle/WQnSNEM3omF7Uruu7 (Google Form linked from the DeepMind announcement).

## Limits

- Drift: visual details and object permanence can shift or drift over time, especially in longer sessions, after multiple complex prompt edits, or while navigating the scene.
- Audio accuracy, legible in-world text, spatial accuracy, and long term consistency are active research areas, not solved problems.
- Real world locations cannot be simulated with perfect accuracy, so a generated street is not evidence about a real street.
- No figures published for resolution, frame rate, maximum session length, benchmarks, price, API rate, or open weights. Any page that fills those cells without a newer Google source is guessing.

## Safety and watermark

- All video and audio generated by Genie 4 carries SynthID, Google's imperceptible watermark, across visual and audio output. It helps identify content as AI generated; it does not prove who made it.
- Do not pay anyone claiming to sell Genie 4 access, an API key, or a download. Google has announced no such product.

## Sources

- Google DeepMind, Genie 4: A new generation of world simulation (October 27, 2026): https://deepmind.google/blog/genie-4-a-new-generation-of-world-simulation/
- TechCrunch, Google's Genie 4 world model adds audio, more realistic worlds, and real-time steering (October 27, 2026): https://techcrunch.com/2026/10/27/googles-genie-4-world-model-adds-audio-more-realistic-worlds-and-real-time-steering/
- madrobot.blog, Genie 4: DeepMind's world model adds sound and talking people (October 27, 2026): https://madrobot.blog/2026/10/27/google-deepmind-genie-4-world-model-sound-dialogue/