• by Colin Kirkland , December 16, 2025

YouTube has announced that it is inviting a select group of creators to use a web app built with Google's newest large language model (LLM), Gemini 3, to help them make small-scale games within the platform's “Playables” ecosystem.

According to an X post from YouTube Gaming, the closed Beta test for “Playables Builder” allows YouTube creators to forgo traditional coding hurdles and develop games with “short text, video or image prompts” managed by Gemini 3.

Specifically, YouTube creators are now able to make new games by either uploading a written description of what they want their game to be, or by uploading a reference image or video to inspire Google’s generative-AI output.

Games developed within the Builder will be immediately playable for users in the U.S., the U.K., Canada and Australia, listed in YouTube's designated “Playables” hub.

YouTube's mini-games launched to all users in May 2024 with over 75 individual titles, including older games that originally premiered on Google HTML minigames service, GameSnacks.

Last winter, YouTube introduced multiplayer capability within Playables, allowing users to play games in real-time with other users, introducing a sense of competition that aligns with gaming content popular within YouTube's video-sharing ecosystem.

Bringing AI into the mix could help expand Playables offerings, as a wider swath of creators without coding knowledge and experience will theoretically be able to transform their ideas for a mini-game into reality, while also further training Google's large-language model in its ability to develop a playable game.