Generate 30-second native videos in one seamless process with up to 50 multimodal references—including images, audio, and video—for consistent characters and scenes. Make precise local edits without needing to regenerate the entire scene.