What the prompts above have in common, and what their clips taught us. Starting from a photo rather than a blank frame? Motion writing for a still is covered on Kling 4.0 image to video.
Name the shot before the subject
Nine of the twelve prompts open with the framing: "Macro close-up", "Wide shot at dusk", "Handheld follow shot". That phrase sets distance and scale before anything else gets invented. Then come the subject and what it does, one camera move, the light and look, and sound last. Keeping that order pays off when a clip goes wrong, because you know which sentence to rewrite. The paper fox opens with its style instead, since there the style matters more than the framing.
Write sound as sources, not a mood
Kling 4.0 makes audio for each clip whether you describe it or not, so finish with a "Sound:" sentence naming two or three things a microphone would pick up: "espresso machine hiss and café chatter", "the bicycle chain ticking". A word like "atmospheric" or "cozy" hands the choice back to the model.
Quote dialogue, name the speaker and the language
Put each line in double quotes, tie it to someone in frame and name the language, the way the barista prompt puts "says in English" right before her question. A voice that should stay off camera gets "off-screen", like her customer. Budget about a dozen spoken words for five seconds, fewer when there is action too; the wok cook gets one. Kling’s guide lists nine speech languages for Kling 4.0, with German, French and Hindi the newest (full Kling 4.0 specs).
Time multi-shot clips yourself
Number the shots, give each a time range and its own framing, and set the clip length to the total. The station prompt marks "Shot 1 (0-5s)" and "Shot 2 (5-10s)" and runs 10 seconds; its cut lands right at five. It works the other way too: the dumpling and wok clips each cut to a new angle without being asked, so write "one continuous shot" when you need a single take.
Let the aspect ratio shape the words
The ratio is a generator setting, not part of the prompt, but the description should suit the frame you pick. The 21:9 glacier prompt describes a wide sweep of water, ice and mountains. The 9:16 wok prompt stacks one cook and one wok in a tall frame and even says "vertical". The 1:1 mug prompt gives the square a single object to centre. Kling4 offers 16:9, 9:16, 1:1 and 21:9.
Leave out what the camera cannot see
Feelings and backstory only reach the screen through something visible, which is why the station prompt asks for wet eyes and a palm on the glass rather than "heartbroken". Treat exact numbers as hints: the mug’s 45-degree orbit came out as a small drift. And keep exclusions to the one that would spoil the shot, such as "no text on screen" for a product clip, instead of a long list of things to avoid.