Procedural Content Generation¶
Text, Narrative & Design
Open Source¶
-
Agentic PCG Discovered by AI
2026
An LLM agent system that uses structured tool calling to optimize and edit game levels through iterative evaluation. It supports multiple classic game domains including Zelda, Sokoban, Lode Runner, and Super Mario Bros.
agentlevel-generationllmpcg -
AutoUE Discovered by AI
2026
An end-to-end multi-agent system that automatically generates complete, playable 3D games in Unreal Engine from natural language descriptions. It coordinates specialized agents for asset creation, scene construction, gameplay code synthesis, and automated play-testing while addressing LLM hallucinations through retrieval-augmented generation.
game-generationmulti-agentpcgunreal-engine -
High Dimensional Procedural Content Generation Discovered by AI
2026
Introduces High-Dimensional Procedural Content Generation (HDPCG), a framework that elevates non-geometric gameplay dimensions, such as gravity inversion, parallel-world switching, and temporal dynamics, to first-class coordinates of a joint state space.
hdpcglevel-generationpcg -
Learning Local Constraints for Reinforcement-Learned Content Generators Discovered by AI
2026
A framework that combines constraint-based generators like Wave Function Collapse with reinforcement learning to guarantee both local visual satisfaction and global playability in procedurally generated game levels.
pcgpcgrlreinforcement-learningwfc -
Multiverse: Language-Conditioned Multi-Game Level Blending via Shared Representation Discovered by AI
2026
A language-conditioned multi-game level generator that enables cross-game level blending and generation from natural language descriptions using CLIP and VQ-VAE.
level-generationmulti-gamepcgtext-to-level -
Procedural Content Metageneration via Program Search and Continual Abstraction Discovery Discovered by AI
2026
Proposes a method to evolve complete Python-based level generators using Large Language Models (LLMs) through program search and Continual Abstraction Discovery (CAD), which extracts reusable primitives from high-fitness programs.
llmpcgprogram-search -
A Database-Driven Framework for 3D Level Generation with LLMs Discovered by AI
2025
Presents a multi-phase pipeline for 3D game level generation that utilizes LLMs to construct reusable databases of architectural components and gameplay mechanics, combined with constraint-based optimization to ensure spatial coherence and navigability.
3d-levellevel-generationllmpcg -
BesiegeField: Agentic Design of Compositional Machines Discovered by AI
2025
An interactive physics-based sandbox and benchmark built on the game Besiege, designed to evaluate and train LLM agents in compositional machine design, assembly, and iterative optimization.
benchmarkllmpcgphysics -
BuildingBlock Discovered by AI
2025
A hybrid 3D building generation framework that integrates generative models, procedural content generation (PCG), and large language models (LLMs) to produce diverse, structured, and hierarchically coherent buildings.
3d-generationllmpcg -
Evolutionary Level Repair Discovered by AI
2025
A search-based framework using evolutionary and quality-diversity algorithms to automatically repair non-functional game levels, ensuring playability and reachability while retaining designer intent.
evolutionary-algorithmlevel-repairpcgplayability -
IPCGRL: Language-Instructed Reinforcement Learning for Procedural Level Generation Discovered by AI
2025
An instruction-based procedural content generation framework via reinforcement learning that incorporates natural language instructions to guide level generation.
language-instructedpcgpcgrlreinforcement-learning -
MarioDiffusion Discovered by AI
2025
A text-to-level generation framework using diffusion models and various text encoders to generate playable tile-based game levels for Super Mario Bros from descriptive captions.
diffusionlevel-generationpcg -
PCGRLLM Discovered by AI
2025
A framework that leverages Large Language Models (LLMs) to automatically design and iteratively refine reward functions for reinforcement learning agents tasked with procedural level generation.
llmpcgpcgrlreinforcement-learning -
The Procedural Content Generation Benchmark Discovered by AI
2025
An open-source benchmark and evaluation framework designed to test and compare procedural content generation (PCG) algorithms across 12 diverse game-related problems, including level generation and rule-set creation.
benchmarkevaluationlevel-generationpcg -
UnrealLLM Discovered by AI
2025
A multi-agent framework that connects natural language descriptions with Unreal Engine 5's Procedural Content Generation (PCG) system. It translates text into executable PCG blueprints and utilizes a spline-based control mechanism for geometric arrangement.
3d-scenellmpcgunreal-engine -
Video Game Level Design as a Multi-Agent Reinforcement Learning Problem Discovered by AI
2025
A multi-agent reinforcement learning framework for procedural level generation (PCGRL) that addresses the computational bottlenecks of single-agent generators by allowing multiple agents to collaboratively design levels.
level-generationmarlpcgpcgrl -
Word2Minecraft Discovered by AI
2025
A system leveraging large language models to generate playable 3D game levels in Minecraft based on structured narrative stories, mapping story elements to gameplay challenges.
level-generationllmminecraftpcg -
Zero-shot 3D Map Generation with LLM Agents Discovered by AI
2025
A training-free dual-agent architecture that leverages LLM agents for zero-shot PCG parameter configuration. It uses an Orchestrator, an Actor (creative generator), and a Critic (quality evaluator) to bridge the semantic gap between abstract user instructions and complex, opaque PCG parameters.
3d-mapllm-agentsparameter-tuningpcg -
3D Building Generation in Minecraft via Large Language Models Discovered by AI
2024
A framework that leverages Large Language Models to procedurally generate complex 3D buildings in Minecraft from natural language prompts, utilizing an intermediate representation and a repair mechanism.
3d-generationllmminecraftpcg -
ChatPCG: Large Language Model-Driven Reward Design for Procedural Content Generation Discovered by AI
2024
A framework that leverages large language models to automatically design and optimize reward functions for reinforcement learning-based procedural content generation.
llmpcgpcgrlreward-design -
Consistent Game Content Creation via Function Calling for Large Language Models Discovered by AI
2024
A framework that leverages LLM function calling to ensure structural consistency and adherence to design constraints when procedurally generating game levels and content.
function-callinglevel-generationllmpcg -
DI-PCG Discovered by AI
2024
A diffusion-based inverse procedural content generation framework that automatically finds optimal parameters for 3D procedural generators to match target shapes or conditions.
3d-assetdiffusionpcg -
DreamGarden: A Designer Assistant for Growing Games from a Single Prompt Discovered by AI
2024
An LLM-driven planning and design assistant for Unreal Engine that recursively refines high-level user prompts into actionable design plans and task sequences to grow diverse game environments.
game-designllmpcgunreal-engine -
Game Generation via Large Language Models Discovered by AI
2024
An LLM-based framework that simultaneously generates both game rules and levels using Video Game Description Language (VGDL), expanding procedural content generation to complete game design.
game-designllmpcgvgdl -
PCG Extended Toolkit Discovered by AI
2024
An open-source plugin for Unreal Engine's Procedural Content Generation (PCG) framework, adding advanced graph theory, pathfinding, spatial operations, and asset management tools.
graph-theorypcgpluginunreal-engine -
Procedural Content Generation in Games: A Survey with Insights on Emerging LLM Integration Discovered by AI
2024
A comprehensive survey of procedural content generation (PCG) in games, focusing on recent deep learning advances and the integration of Large Language Models (LLMs). It analyzes how LLMs are disrupting traditional PCG pipelines and enabling more sophisticated, controllable, and interactive game content creation.
deep-learningllmpcgsurvey -
Procedural Content Generation via Generative Artificial Intelligence Discovered by AI
2024
A comprehensive survey examining how modern generative AI techniques including diffusion, LLMs, and GANs are applied to procedural content generation for game terrains, dungeons, items, and levels.
generative-ailevel-generationpcgsurvey -
Word2World Discovered by AI
2024
An LLM-based procedural content generation system that designs playable 2D game worlds and narratives directly from stories without task-specific fine-tuning.
level-generationllmpcg -
Godot 4 Procedural Generation Discovered by AI
2023
A collection of procedural content generation (PCG) algorithms and interactive demos implemented for the Godot 4 game engine. It includes implementations of popular techniques like Random Walker (as seen in Spelunky), cellular automata, and shader-based world map generation with biomes and rivers.
algorithmsgodotlevel-generationpcg -
Level Generation Through Large Language Models Discovered by AI
2023
A research project exploring the capability of Large Language Models to generate functional, playable Sokoban game levels by training on textual representations of level layouts.
level-generationllmpcgsokoban -
MarioGPT Discovered by AI
2023
An open-source text-to-level generation model that uses a fine-tuned GPT-2 architecture to generate Super Mario Bros levels from natural language descriptions and control parameters.
gpt-2level-generationllmpcg -
RaidEnv: Exploring New Challenges in Automated Content Balancing for Boss Raid Games Discovered by AI
2023
Presents RaidEnv, a reinforcement learning environment designed for automated content balancing and difficulty adjustment in boss raid games, reducing manual labor for game designers.
content-balancingpcgreinforcement-learning -
Reconstructing Existing Levels through Level Inpainting Discovered by AI
2023
A procedural content generation framework that introduces level inpainting to reconstruct and extend video game levels. It adapts autoencoder and generative adversarial network techniques from image inpainting to the game level domain.
level-generationpcgpcgml -
Illuminating Diverse Neural Cellular Automata for Level Generation Discovered by AI
2022
A quality diversity (QD) approach to training diverse neural cellular automata (NCA) level generators whose output levels vary based on aesthetic or functional criteria.
level-generationncapcgquality-diversity -
Edgar-Unity Discovered by AI
2020
A graph-based procedural level generator for Unity that allows developers to generate 2D dungeons and levels with complete control over the layout. It uses a constraint-based approach to place hand-designed room templates according to a specified level graph.
dungeon-generatorlevel-generationpcgunity -
PCGRL: Procedural Content Generation via Reinforcement Learning Discovered by AI
2020
A foundational framework and OpenAI Gym environment that frames procedural content generation as a sequential decision-making task solved via reinforcement learning.
gympcgpcgrlreinforcement-learning -
WaveFunctionCollapse (WFC) Discovered by AI
2016
A highly influential procedural generation algorithm that generates bitmaps or tilemaps from a single example using constraint satisfaction inspired by quantum mechanics.
pcgtilemap-generationwfc
Proprietary / Industry Tools¶
-
Summer Engine Discovered by AI
2025
An AI-driven game engine compatible with Godot that allows creators to build 2D and 3D game levels, scenes, and mechanics through natural language conversations.
game-enginelevel-generationpcg -
Unreal Engine PCG Framework Discovered by AI
2023
Epic Games' built-in toolset for creating procedural content and tools directly inside Unreal Engine, enabling artists and designers to build fast, iterative environments and biomes.
biome-generationpcgunreal-engineworld-building -
SideFX Houdini Discovered by AI
1996
An industry-standard 3D procedural software widely used in game and film production for modeling, terrain generation, VFX, and creating reusable procedural assets via Houdini Engine.
pcgprocedural-modelingterrain-generationvfx
