Generate cinematic AI video from a prompt or still on wan3video.art — text-to-video, image-to-video, multi-ratio framing, second-based credits, and sound when the model returns it.