@matrix watching 2kliksphilip's video on this, that video of the car driving with the camera following behind it is a track ripped from a video game just with the theme and vehicle changed. i can't help but wonder how much of this was generated by itself after training on video content and how much of it was done with a technique similar to image2image, but with video2video. as in, they took footage of the game and gave the AI the prompt and that's why it looks as smooth and good as it is. perhaps the ones that look much less canny are the ones that had nothing to go off but the prompt.