Genie 4 Gemini API: Interactive AI Worlds Open to Developers
Google opened the Genie 4 Gemini API and AI Studio access. What the world model does, what it costs per second, and how marketers can test it safely.

In this article
The Genie 4 Gemini API launch puts Google’s world model in developers’ hands. Google announced it in a developer post on November 19, 2026, the day after Genie 4 reached Project Genie. Developers can call the model through the Interactions API in the Gemini API or try it in Google AI Studio.
Genie 4 is a world model. It does not make a fixed video clip. It makes a scene you can move through with a keyboard, and it keeps generating as you go. For marketers it is still an early tool, but you can now build it into your own apps and prototypes.

What Genie 4 in the Gemini API can do
Google’s post lists these capabilities:
- Continuous 720p worlds at 24 frames per second, with synchronized native audio such as ambient sound, sound effects and character dialogue.
- A start from a text prompt, a reference image, or both.
- Real time keyboard controls, plus live text prompts that change the weather, lighting, objects or events while you explore.
- Gemini reasoning to keep the state of the world consistent.
- A SynthID watermark on every output.
Google names three uses for developers: prototyping interactive experiences, visualizing spatial designs, and training robots and AI agents. PPC Land points out that advertising is not on that list.
How we got here
Genie 4 is new. TechCrunch reported on October 27, 2026 that Google announced it as a research preview for trusted testers. TechCrunch said it can keep a consistent scene for up to two minutes, double the limit of the model before it.
| Date | What happened | Source |
|---|---|---|
| October 27, 2026 | Genie 4 announced as a research preview for trusted testers | TechCrunch |
| November 18, 2026 | Genie 4 launched in Project Genie for eligible Google AI Ultra subscribers | Google, PPC Land |
| November 19, 2026 | Genie 4 available in the Gemini API and Google AI Studio |
Genie 4 Gemini API pricing and limits
Google’s post links to a pricing page but does not state the price itself. Trade reports fill in some detail.
PPC Land reports a cost of “approximately $0.10 per second of streamed output, stereo audio included,” a session cap of 120 seconds, and the model name models/genie-4. It also says you need a paid tier Gemini API key. Unite.AI reports that Genie 4 is not offered on the free tier and gives a video output price of $17.50 per million tokens.
At about $0.10 per second, one full 120 second session costs roughly $12. Unite.AI also warns that closing a browser tab does not immediately stop generation on the server, so a careless prototype can keep billing. Check Google’s own pricing page before you budget, because these figures come from trade reports, not from the announcement.
Where Genie 4 could help marketing teams
In our view, the model suits early, internal work that fits Google’s own use cases better than customer facing campaigns.
- Prototyping an interactive experience, such as a walk through a store, a booth or a product launch space, before paying a 3D studio.
- Visualizing a spatial design, such as an event layout, to share with a client.
- Testing agent ideas, since Google lists training AI agents as a use case.
If you are already using Gemini for content work, our guides to Gemini Notebook interactive sites and Gemini Omni 1.2 Flash video cover tools that are closer to everyday marketing output.
What marketers should do now

These are our recommendations, not official Google guidance.
- Try it in Google AI Studio first. Google points developers to an interactive canvas and starter apps there, so you can judge quality before writing code.
- Set a per session budget. Work out the cost of the sessions you plan, using the price on Google’s pricing page.
- End sessions on purpose. Build a clear stop into any prototype so a closed tab does not keep a session running.
- Keep the SynthID watermark in mind. Every output is marked as AI made, so plan to label any public use as AI generated.
- Check your ad platform rules before using clips in ads. Google Ads already shows AI labels in some regions, as we covered in Google Ads AI labels.
- Write down what you learn. Treat early tests as research and share the results with your team before you commit spend.
For a wider view of Google’s newest models, see our explainer on Gemini 4 Neon and our guide to the Gemini Decisions API.
FAQ about the Genie 4 Gemini API
When did Genie 4 become available in the Gemini API?
Google announced it on November 19, 2026. It is available through the Interactions API in the Gemini API and in Google AI Studio.
How much does Genie 4 cost?
Google’s post does not state a price in the text. PPC Land reports about $0.10 per second of streamed output, and Unite.AI says it is not on the free tier.
How long can a Genie 4 session last?
PPC Land and Unite.AI both report a cap of 120 seconds per session. TechCrunch said in October that Genie 4 can keep a scene consistent for up to two minutes.
Can I use Genie 4 output in ads?
Google does not list advertising as a use case. Every output carries a SynthID watermark, so label any public use as AI generated and check the rules of the ad platform you plan to use.




