Hot AI News

Developers can now build on Genie 4’s 720p worlds with sound through the Gemini API

A day after Project Genie got Genie 4, Google opened the world model to developers through the Interactions API and AI Studio, with 720p video, native audio and SynthID watermarks.

Developers can now build on Genie 4’s 720p worlds with sound through the Gemini API
Image: Google

Google has opened Genie 4, its interactive world model, to developers through the Gemini API and Google AI Studio. According to the official announcement, it generates continuous 720p worlds at 24fps with synchronized audio.

The release comes one day after Genie 4 reached consumers in Project Genie. Our earlier report on Project Genie covered that side; this one is for people who want to build on the model.

Developers can now build on Genie 4's 720p worlds with sound through the Gemini API
Image: Google

What developers can generate and control

Access runs through the Interactions API, with AI Studio as the no-code entry point. A world can start from a text prompt, a reference image, or both. Google lists the audio as ambient soundscapes, sound effects and character dialogue, all generated natively alongside the video.

Once a world is running, the user moves through it with keyboard controls and can type live prompts to change weather, lighting, objects or events. Google says Gemini reasoning keeps the world state consistent as those changes come in. The company names three uses: prototyping interactive experiences, visualizing spatial designs, and training robotics and AI agents.

Developers can now build on Genie 4's 720p worlds with sound through the Gemini API
Image: Google

Every output carries imperceptible SynthID watermarking, Google says. AI Studio offers an interactive canvas and starter apps, and the announcement sends readers to the Gemini API documentation for pricing.

Advertisements

Open details that the announcement leaves out

The post gives no price, no session length, no rate limits and no list of supported regions. It also does not say whether access needs a paid tier or a waitlist. Anyone planning a project has to read the pricing page before committing, because those numbers decide whether continuous generation is practical outside a demo.

Developers can now build on Genie 4's 720p worlds with sound through the Gemini API
Image: Google

The performance figures are Google’s own description. The post includes no benchmark and no comparison with Genie 3, and no independent testing has been published in the sources we reviewed.

A research model moving to developers within weeks

The speed of the rollout is the clearest number in the story. When Genie 4 was first shown on October 27, it was limited to trusted testers, as we reported in our coverage of that preview. Google said then that wider access would come through Project Genie for AI Ultra subscribers and through early-access partners in the Gemini API and AI Studio. Three weeks later, the API is open to developers.

The consumer product moved on a narrower track. Project Genie went to eligible AI Ultra subscribers, while AI Pro members were offered a waitlist, according to our earlier report. Developers using the API do not appear to face the same plan gate, although the announcement does not say so directly.

The limits Google disclosed at the October preview still matter for anyone building on the model. Google listed visual drift, objects that lose permanence over long sessions, and weak text rendering, spatial accuracy and physical interactions. A DeepMind engineer told TechCrunch it is “still early days for Genie 4.” Nothing in the API announcement says those limits have been fixed.

The robotics and agent-training use case is the most ambitious one in the post. A world model is only useful as a training environment if its physics hold up, and Google’s own list of physical-interaction weaknesses suggests this is unproven. For design visualization or game prototyping, visual coherence matters more, and 720p at 24fps with audio is a reasonable bar for early experiments.

The release follows the launch of Gemini 4 Neon a day earlier, whose pricing we covered in our Gemini 4 Neon report. Google is shipping the Gemini API’s newest models in quick succession, and Genie 4 now sits in the same developer surface as its text models.

Sign Up for the Daily AI Pulse

One email a day. All the stories that matter.

Alex CarterAI News & Big Tech Correspondent

Alex Carter writes for Aiholics, keeping readers updated on the fast-paced world of AI and Big Tech.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may also like