Adaptive LLM-based Game Radio (ALGR)
Paul Mieschke, Stefan Radicke · 2023
This paper, Adaptive LLM-based Game Radio (ALGR), proposes an implementation for a dynamic in-game radio system, that is able to create infinite (non-music) content influenced by the game state with the help of Large Language Models (LLMs) and Text-To-Speech (TTS) technology. The radio is provided with information about the game by a general context and game context, which it can reference in its content. Additionally, it is rooted in the game world as a diegetic sound system. Both of these facts further increase player immersion.Overall, convincing results have been achieved based on rudimentary evaluation.