Chinese AI startup JIYI SOON has launched SOON, a conversational game-creation platform that aims to...
The AMW Read
Another CN entrant in AI game generation with no pricing, funding, or named partners; updates the generative-media player map but does not shift structural forces.
Chinese AI startup JIYI SOON has launched SOON, a conversational game-creation platform that aims to compress early-stage game development into a text-to-playable workflow.
SOON lets users describe a game in natural language and has AI handle code generation, game logic, characters, scenes, and UI. Rather than simply simplifying an existing engine — the path taken by Paris-based no-code engine GDevelop, which began as an open-source project in 2008 and incorporated in 2021 — SOON tries to reduce the manual work a creator performs inside that workflow. The team has made a deliberate trade-off: the platform currently focuses on 2D and mini-game projects, a narrower scope than GDevelop's 2D and 3D support, but one that lets SOON automate more of production while keeping output relatively stable. In its second round of testing, users created more than 3,000 game projects in 45 days, with the platform generating code, game logic, characters, scenes, and UI.
this is a bet on the creator-economy end of generative media, not the AAA pipeline. Large commercial games still depend on experienced teams, precise control, and sophisticated development environments, so the near-term addressable market is prototypes, small projects, and creator-driven games. The strategic signal is the Chinese ecosystem's willingness to push generative tooling into content-creation verticals — gaming first, with distribution reach through consumer platforms — rather than competing head-on with frontier model labs on general-purpose capability. It also echoes a well-worn accessibility arc: Unity and Unreal gave small teams pro-grade tools, no-code engines removed the programming barrier, and the current generation is trying to remove the engine-learning step entirely.
For builders and investors: the constraint to watch is output stability at the constrained 2D/mini-game scope, not feature breadth. If conversational generation reliably produces shippable prototypes, the moat question shifts from model quality to distribution — which storefront, engine, or platform surface owns the creator relationship. A platform that only generates games but does not own publishing will find its output commoditized as similar capabilities land inside incumbent engines.