Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)

Google opens Genie 4 world model in its API: one minute of walking costs about $6

Google has opened Genie 4, a "world model," through its API. From a prompt or an image, it builds a world you can walk through, with sound and character voices. The picture runs at 720p and 24 frames per second, and you can change the weather, the lighting, or add a new character with text in the middle of a walk. A minute costs about $6 and an hour roughly $360, so for now it is a tool for prototyping games and AR/VR, not for mass-market products.

AI-processed from TG: @TochkiNadAI (Точки над ИИ); edited by Hamidun News
Google opens Genie 4 world model in its API: one minute of walking costs about $6
Source: TG: @TochkiNadAI (Точки над ИИ). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTG: @TochkiNadAI (Точки над ИИ) ↗
Editor-in-Chief
720p and 24 fpssound is built in, and you can talk to NPCs by voice
Priceabout $6 per minute and roughly $360 per hour of walking
Key metric #3Google AI Studio is currently giving away credits for testing — look for the gift in your profile

Until now, Google's Genie world model could only be watched in demo videos. Genie 4 is now available through an API. That means you can order an interactive world from your own app the same way you order text or an image.

Imagine you are building a game or an AR/VR project. Instead of spending weeks assembling a test location, you type "night city in the rain" into a prompt or upload a screenshot, and the model produces a space you can walk through. It runs at 720p (HD picture) and 24 frames per second, the same as film, and costs about $6 per minute and roughly $360 per hour.

This comes from the author of the Telegram channel "Точки над ИИ" ("Dots over AI"), who tried the model himself. It is worth reading for anyone working on games, AR/VR, and simulators: it makes it easier to see what this technology can do today and what it costs.

What Genie 4 can do

The principle is simple. You write a prompt or drop in an image, and Genie builds a world you can walk through. It is not a pre-recorded clip: the image is generated as you move. Sound comes built in, from background noise to characters' lines.

The most interesting part is control on the fly. During the walk itself, you can use text to change the weather or the lighting, or drop a new character into the scene. According to the author, the world holds together and does not fall apart when you do this. You can talk to characters by voice: the author's own voice did not record, but he still managed to talk with NPCs (non-player characters).

As a test, he asked the model to generate a world in the spirit of GTA 6. In his observation, the model showed no copyright restrictions, and his channel's logo appeared on a billboard in the scene. This is one user's experience, not official Google policy. If you are considering commercial use, read the legal terms separately.

How much it costs

The prices the author cites are about $6 per minute of generation and roughly $360 per hour of walking. Converted, that comes to about 10 cents per second. For a consumer game, where a player spends hours in a world, that price does not work yet. For short tasks, though, it looks acceptable:

  • a quick prototype of a level or location to show an idea to a team or an investor;
  • AR/VR concept validation before hiring artists and engineers;
  • short demos for pitches and conferences;
  • experiments with NPCs that react to voice and text.

Limitations and what to do

The author has 12 years of experience in AR/VR, and by his own account the very idea of a "whole interactive world via API" makes his fingers itch to build. But he names the weak spots plainly. First, he wants 120 frames per second: 24 frames is too few for VR, because at a low frame rate people quickly get motion sick. Second, he would like fewer model hallucinations, meaning fewer cases where details of the world "drift" or are made up. By his estimate, it is about a year until the model reaches that level. This is the author's personal estimate, not a promise from Google.

Practical takeaways for developers and product managers:

  1. Do not build your core game loop on Genie 4. Use it as a prototyping tool.
  2. Set a budget in advance: at $6 per minute, costs add up fast if you test a lot and without limits.
  3. Check the legal side if you plan to publish the output, especially with familiar brands and franchises.
  4. Watch the price and the frame rate. Once they approach the values you need, interactive worlds via API will become a real building block of a product.

You can try it right now: according to the author, Google AI Studio is currently giving away credits. In the left corner, where the profile is, there should be a small window with a gift. The post does not specify the terms of the promotion, so it is better to check the details directly in the interface. It is the cheapest way to evaluate quality and price before launching a project.

Story

Key Questions Answered

What Genie 4 can do?

The principle is simple. You write a prompt or drop in an image, and Genie builds a world you can walk through. It is not a pre-recorded clip: the image is generated as you move. Sound comes built in, from background noise to characters' lines. The most interesting part is control on the fly. During the walk itself, you can use text to change the weather or the lighting, or drop a new character into the scene. According to the author, the world holds together and does not fall apart when you do this.

How much it costs?

The prices the author cites are about $6 per minute of generation and roughly $360 per hour of walking. Converted, that comes to about 10 cents per second. For a consumer game, where a player spends hours in a world, that price does not work yet.

What are the key details on Limitations and what to do?

The author has 12 years of experience in AR/VR, and by his own account the very idea of a "whole interactive world via API" makes his fingers itch to build. But he names the weak spots plainly. First, he wants 120 frames per second: 24 frames is too few for VR, because at a low frame rate people quickly get motion sick. Second, he would like fewer model hallucinations, meaning fewer cases where details of the world "drift" or are made up. By his estimate, it is about a year until the model reaches that level.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Private AI Builders Community · Vibe Club

Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews

Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.

Join Vibe Club →Explore Hamidun Academy
What do you think?
Loading comments…