Supplementary videos: zero-shot generation from real-video first frames

Submission 18051, rebuttal material. The model sees only the first frame (a real photo from a handheld capture) and a camera path taken from that capture (translations rescaled to the training speed, intrinsics fixed at 90° horizontal FOV). The model was trained on synthetic game-engine video only; nothing here was tuned on real video. The real clip on the left follows the same path and is shown for reference; the two are not expected to match pixel for pixel.