From LLM-Driven Trading Card Generation to Procedural Relatedness: A Pokémon Case Study

2026-04-30

AI Summary

Traditional collectible card games often suffer from strategic homogenization and constrained card selection due to rigid metagame structures, diminishing player engagement. This work proposes a generative framework that integrates player co-creation, fine-tuned embeddings, locally deployed large language models, and image diffusion models to introduce procedural relatedness into card design for the first time, enabling dynamic co-generation of game mechanics and visual styles. The approach empowers players to express creative intent efficiently through prompt engineering while ensuring privacy and interactive responsiveness via local deployment. A user study (N=49, generating 196 cards) demonstrates high participant satisfaction with both mechanical coherence and visual quality, validating the framework’s effectiveness in expanding creative boundaries, fostering emotional connections, and transcending conventional metagame evolution patterns.

Abstract

Since the dawn of Trading Card Games, the genre has grown into a multi-billion-dollar industry engaging millions of analog and digital players worldwide. Popular TCGs rely on regular updates, balance adjustments, and rotating constraints to sustain engagement. Yet, as metagames stabilize, predictable strategies dominate and viable card options diminish, often resulting in repetitive and impaired player experiences. This paper investigates the use of Large Language Models and Image Diffusion Models for Procedural Content Generation of TCG cards, addressing these challenges by enabling a personalized infinity of card designs. Modern generative AI not only enables large-scale content creation but could even introduce procedural relatedness, fostering unique connections between players and their cards. We present a pipeline combining player-centric co-creation, fine-tuned embeddings, local LLMs, and Diffusion Models to generate dynamic, personalized cards while potentially expanding creative range. We evaluated the pipeline in a user study with 49 participants who generated 196 Pokémon card samples. Participants rated aesthetics and representativeness of visuals and mechanics, and provided qualitative feedback. Results show high satisfaction and indicate that most participants successfully realized their own ideas through prompt adjustments. These findings lay groundwork for future content generation systems and alternatives to conventional metagame evolution through procedural relatedness.

Problem

• Trading Card Games
• metagame stagnation
• procedural content generation
• player experience
• card diversity

Innovation

• Procedural Content Generation
• Large Language Models
• Diffusion Models
• Procedural Relatedness
• Player-Centric Co-Creation