Genie 4 Opens to Developers at $0.10 a Second
Google on Nov. 19 made Genie 4 available to developers through the Gemini API and Google AI Studio, a day after its AI Ultra rollout. Paid-tier keys stream 720p, 24 fps worlds with stereo audio at about $0.10 per second, capped at 120 seconds per session.
Google on Nov. 19 opened Genie 4, its real-time interactive world model, to developers through the Gemini API and Google AI Studio, one day after rolling it out to eligible Google AI Ultra subscribers. Google's developer documentation puts the price at roughly $0.10 for every second of streamed output, with each session capped at 120 seconds.

Key takeaways
- Genie 4 (model code models/genie-4) is now available on paid-tier Gemini API keys through the Interactions API, plus an interactive canvas in Google AI Studio.
- Output is 1280x720 video at 24 frames per second with 48 kHz stereo audio, billed at 240 video output tokens per frame, or about $0.10 a second.
- Sessions stop at 120 seconds, cannot be resumed, and may lose visual stability after 60 to 90 seconds, according to Google.
What developers get with the Genie 4 API
In a short post on The Keyword, Google said developers can start from a text prompt, a reference image or both to generate "continuous 720p worlds at 24fps with synchronized native audio," ranging from ambient sound and effects to character dialogue. Keyboard controls move the viewer, while live text prompts change "weather, lighting, objects, or events on the fly."
Google named three intended uses: prototyping interactive experiences, visualizing designs for spatial concepts, and training robotics and AI agents. Every output carries Google's invisible SynthID watermark, the company said.
Genie 4 pricing and technical limits
The Genie 4 model page lists the model as stable and requires a paid-tier key. Video is billed at 240 output tokens per frame, or 5,760 tokens per second, with audio bundled in. By AI Newsway's arithmetic at Google's approximate rate, one full 120-second session runs 2,880 frames and about $12. When AI Newsway checked the Gemini API pricing page on Nov. 20, it did not yet carry a separate Genie 4 entry.
Batch, Flex and Priority inference are not offered. The model also lacks function calling, Google Search grounding and structured outputs.
Google's integration guide says the first frame takes roughly 5 to 9 seconds to arrive. Movement keys respond in about 0.2 to 0.5 seconds, while text prompts take about 1 to 2 seconds to show up. Browser apps connect over WebRTC and Python or Node.js code over WebSocket, using single-session tokens so the main API key stays on the server.
The guide warns that only calling complete() stops billing; simply disconnecting does not. Model serving runs from data centers in the United States, Western Europe and Japan, with round trips under 150 milliseconds in North America, Western and Central Europe, and East Asia.
Known limitations
Genie 4 keeps a rolling visual window rather than a persistent 3D scene. Looking away from an object for eight seconds or more, or turning a full 360 degrees, can make landmarks shift or reset. Output is fixed at 16:9, with no portrait 9:16 option.
Prompts or reference images containing identifiable real people, public figures or copyrighted brand IP "can be blocked at startup or terminate an active stream mid-session," the guide says.
Context: from research preview to paid API
Google DeepMind introduced Genie 4 on Oct. 27 as a research preview for trusted testers, without prices or specifications. On Nov. 18, Google brought it to eligible AI Ultra subscribers in Project Genie, and AI Newsway reported then that no API date had been given. The developer release came the next day, 23 days after the first announcement.
What's next
Google has not published a per-token rate for Genie 4 or limits on how many sessions a project can stream at once. It also has not said when sessions might run longer than two minutes.
FAQ
Can free Gemini API users try Genie 4?
No. The API requires a paid-tier key. Developers can also try it through Google AI Studio's interactive canvas and starter apps.
How much does a Genie 4 session cost?
Google lists about $0.10 per second of streamed output. A maximum 120-second session works out to roughly $12 at that rate.
Can I make vertical videos for social media?
Not yet. Genie 4 outputs only 16:9 landscape video at a fixed 24 frames per second.
References
- Genie 4 is now available on the Gemini API. — Google, The Keyword (Nov. 19, 2026)
- Genie 4 — Google AI Studio documentation (Nov. 2026)
- Interactive world generation with Genie 4 — Google AI Studio documentation (Nov. 2026)
- Genie 4 is now available — Google, The Keyword (Nov. 18, 2026)
Editor responsible for AI Newsway. Writes about AI, developer tools, cloud and productivity technology.
Comments 0
Comments are currently disabled.