I’m not the author, but from what I know it is completely generative, in that, some kind of ml model directs the scene and audio. My intuition is its probably handed off to some kind of interpreter that generates the actual scene with the animations and everything, but again I don’t really know.