Moonshine: Distilling Game Content Generators into Steerable Generative Models

TL;DR

This paper introduces a method to distill procedural game content algorithms into controllable, text-conditioned generative models, improving content diversity, accuracy, and controllability in game design.

Contribution

It presents a novel distillation process that converts constructive algorithms into controllable PCGML models conditioned on plain text, enabling flexible game content generation.

Findings

Distilled models match the original algorithm's output quality.

Text conditioning provides effective controllability.

Models outperform baseline algorithms in diversity and accuracy.

Abstract

Procedural Content Generation via Machine Learning (PCGML) has enhanced game content creation, yet challenges in controllability and limited training data persist. This study addresses these issues by distilling a constructive PCG algorithm into a controllable PCGML model. We first generate a large amount of content with a constructive algorithm and label it using a Large Language Model (LLM). We use these synthetic labels to condition two PCGML models for content-specific generation, a diffusion model and the five-dollar model. This neural network distillation process ensures that the generation aligns with the original algorithm while introducing controllability through plain text. We define this text-conditioned PCGML as a Text-to-game-Map (T2M) task, offering an alternative to prevalent text-to-image multi-modal tasks. We compare our distilled models with the baseline constructive algorithm. Our analysis of the variety, accuracy, and quality of our generation demonstrates the efficacy of distilling constructive methods into controllable text-conditioned PCGML models.