The core principle of writing video prompts is to let the model guess as little as possible—tell it the information it needs upfront—rather than simply making the prompt as long as possible.