They are not pre-rendered voice lines? I had assumed they were using LLMs for asset creation, not live. (It seems way too risky for a AAA studio to release a multiplayer game played by kids where you use a genAI prompt live, but they do keep surprising me by how little they care about brand safety)