I think I kinda understand how Sora works now So Stable Diffusion / DALL-e for example is a 2d model so it has X/Y coords Then Stable Video Diffusion, one of the first video models, is the same but with a temporal dimension, it can predict the next frames, so it has X/Y/t (time) coords Then Sora is one extra dimension, it's a 3d point cloud so X/Y/Z, but then with t (time) temporal dimension added But I have no idea