Google announced Playground, a browser-based platform for creating games from AI prompts, on 7 October 2026, The Verge reported. Available in the US to people aged 18 and older at playground.google, it lets someone describe a game in conversation, try a playable version and ask for changes without writing code. Google is also preparing an integration with Unity Spark, a separate browser-based game creation tool due in beta later in 2026.
Key points
- Playground combines Google’s Gemini, Nano Banana and Lyria models with a custom system for creating games from prompts.
- Games can be shared by link or published to a gallery; some genres support multiplayer and leaderboards.
- Playground is free, with higher weekly token limits tied to Google One subscription plans.
- Unity Spark is planned for beta later in 2026, but projects made with it cannot initially be exported to the full Unity engine or sold.
Gemini, Nano Banana and Lyria in Playground
Playground begins with a chat rather than a conventional game editor. A user can start with an idea, adapt one of Google’s suggested prompts or follow guided help, then request changes to a playable result. The choices include a 2D or 3D game and single-player or multiplayer play, 9to5Google reported. Google says requests can extend from changing characters and scenery to adjusting physics and rules, so creation is presented as a sequence of revisions rather than a single prompt that must get everything right.
Nia Carter, a Google spokesperson, told The Verge that Playground uses the company’s Gemini, Nano Banana and Lyria foundation models alongside a custom harness refined using games and evaluations made internally. The harness is the system that brings the models into the game-making process: a prompt reaches a tool built to produce something playable, rather than an ordinary chat response alone. Carter described the internal games and evaluations as part of Google’s development of that system. The account concerns Google’s own work on Playground, rather than a reported test of games made by its users.
Making a game to share with friends could begin with a description of its rules, continue with requests to change its characters or physics, and end with a link to the playable result. It could be opened in a browser on a phone or laptop, without the person making it having to write the code described in the prompt.
Google describes Playground as an experiment aimed at people without coding experience. That proposition extends beyond generating an image or a piece of text: the result has to respond when played, and later instructions have to alter the experience the creator is testing. Playground’s conversation allows that back-and-forth, while the account of its custom harness places the work of assembling a playable game inside Google’s system.
Playground games can reach the Explore gallery
A finished game can stay private, be shared through a link or be published to the Playground Explore gallery. Games run in a browser on a phone or laptop, and people can rate those in the gallery. Google says ratings and play activity will help determine which games the gallery highlights. Select genres support multiplayer and in-game leaderboards, while the games published to the gallery undergo community-guideline screening and can be reported by users.
The release also puts a limit on how much game-making access a person receives. Playground is free, but Google One subscribers receive higher weekly token limits according to their plan, Carter said. Those limits matter to a tool built around repeated requests: the same conversation used to make a game is also how its creator asks for another version. Access to creation is rolling out in tiers based on Google One membership, Android Authority reported, while eligible users can access the site to play games.
Unity Spark is planned for 2026
Unity Spark will bring another way to build games through Playground when its beta launches later in 2026, GamesIndustry.biz reported. It runs on the Unity engine and is designed for creating and revising 3D games through text prompts. Unity Spark uses a range of generative AI models rather than relying exclusively on Google’s technology. Its connection to the Unity Asset Store would let a request for a game element bring up suggested assets made by artists.
In a demonstration to GamesIndustry.biz, a request for a character’s cape returned choices from the asset store, while a request for a sunrise background produced options for that scene. The agent could also propose a way to switch between lighting settings, and several agents could work at once. Those were actions in a demonstration of Unity Spark, which had not yet entered the planned beta.
Unity Spark projects can be worked on together through a shared link and published on Playground. Its editor will run in desktop browsers, although games made with it can be played in desktop and mobile browsers. At launch, projects cannot be downloaded, moved into the full desktop Unity engine or sold. Unity plans a credit system tied to AI tokens, with some free access and further credits available to buy; the payment details had not been finalised.
Unity opened several other generative AI tools for its game engine to beta testing in May 2026. In August, chief executive Matt Bromberg credited the continuing success of Unity Vector AI with the company’s revenue growth. Unity plans to launch Unity 7 in Q1 2027 with AI tools as a core component, and senior vice-president of product Adam Smith has said developers will not be required to use them.
