Back to AI game news
Tool updates

AI Games open-sources a mobile prompt-to-game creator with offline Neon Comet sample

By GameSlash

The Expo-based AI Games project outlines prompt-based 2D/3D game creation and signed offline play, while clearly separating real Neon Comet gameplay from synthetic UI demo states.

2D3DSingleplayer
AI Games app catalog showing the bundled offline Neon Comet sample
Image: AI Games / mcontre4 — Official documentation capture with synthetic demo data (MIT License)

AI Games is an open-source game-creation project for iOS and Android built with Expo and a self-hosted AI generation backend. The repository describes a prompt-to-game flow spanning moderation, artwork, gameplay code, validation and package signing, but the published demo must not be confused with an independently verified end-to-end AI generation run.

Build from a prompt, but play a bundled example offline

The interface lets users choose 2D, 3D or isometric views before describing a game and requesting future changes. The app's native player is designed to verify signed downloads and asset hashes and keep game progress stored on the device. Actual AI creation requires a funded OpenRouter account through the project's Codex CLI gateway and development infrastructure.

AI Games creation screen with 2D, 3D and isometric options, displaying synthetic credit fixtures
Image: AI Games / mcontre4 — Official documentation capture with synthetic demo data (MIT License)

The repository includes a real deterministic Phaser sample named Neon Comet. Unlike the synthetic generation and credit balances in the documentation UI, Neon Comet is identified as a genuinely playable, bundled offline game. Users can inspect a reproducible gallery capturing its browser runtime without a real AI provider.

Original Neon Comet sample gameplay screenshot from the verified offline browser runtime
Image: AI Games / mcontre4 — Official documentation capture with synthetic demo data (MIT License)

Screenshots distinguish working demo from production claims

The project's screenshot provenance document says its UI shots use synthetic accounts, credits, library states and generation progress. The gameplay capture uses an actual packaged runtime with networking disabled, but does not prove native iOS or Android installation or a production generation pipeline. The source is MIT-licensed. GameSlash reviewed the README and screenshots, not a device build or self-hosted deployment.

Sources: AI Games repository | Screenshot methodology | MIT License

View source