All-in-one video generation model supporting text-to-video, image-to-video, first/last-frame, and omni-modal reference-based generation (image, video, and audio references) with synchronized audio, at up to 30 seconds per clip.