Abstract
Multi-scene navigation is a defining feature of contemporary 3D games, yet authoring it is laborious. Every portal must have consistent endpoints, and interiors need to remain navigable after furnishing. Recent LLM and MLLM scene generators have reduced the cost of single-interior synthesis, but they produce one scene at a time and cannot yield a connected multi-scene world through naive repetition.
We identify three obstacles that single-scene methods leave unsolved: cross-scene consistency, in-scene navigability, and the evaluation of whether a transition actually works. We present MAGIC, a prompt-to-project system that addresses all three. MAGIC is a four-stage pipeline that converts a single natural-language prompt into a runnable multi-scene game project:
- Plans a shared transition-aware intermediate representation;
- Specifies each scene while enforcing portal reachability with a flood-fill validator;
- Generates the scenes along with their transition scripts and combines them into one project.
We further introduce a transition-focused evaluation agent that runs each transition in play. On a benchmark of 100 multi-scene cases, MAGIC produces an executable project for every case and achieves 0.99 precision, 0.95 recall, and 0.96 F1 on end-to-end transition identification. Stage by stage, it recovers more ground-truth portals and yields markedly more navigable layouts than an LLM baseline and Holodeck.
Our code is available at GitHub.
Blogger's Review: The MAGIC project illustrates how an innovative four-stage pipeline can tackle the complexities of multi-scene game world generation. Its focus on transition evaluation mechanisms not only enhances generation efficiency but also alleviates burdens on game designers, advancing the automation of game creation.