Writing a shot list for Kling 4.0's ten keyframes

A screenwriter already writes the thing Kling 4.0 asks for. Its new multiple-keyframes feature — up to ten images pinned to moments in a clip of up to 30 seconds — is a shot list with pictures, and Kling's own guide to it, published on 30 September, reads like advice on blocking a scene.

Keyframes are the beats, the prompt is the action

Kling's guide puts the division in one sentence: the keyframes show the important moments, and the prompt explains how the video moves between them. A keyframe is not a shot; it is a state the scene must reach — the door open, the letter read, the face turned. Everything that happens on the way there belongs in the words.

Lay the prompt out as a timeline

Kling 4.0 accepts prompts of up to 8,000 tokens, and the guide recommends spending them on a timeline: one segment per stretch of time, written as 0–5s:, 5–10s: and so on, each saying what happens and how the camera sees it. Its own example runs from 0–5s, a kitchen at sunrise and a woman walking to the counter, into 5–12s, where she picks up a product and the camera moves closer: one segment per change in the situation.

Fewer keyframes than you think

Ten is a ceiling, not a target. The guide's test for whether a moment deserves one is whether anything important changes there; if nothing does, it says, you probably do not need another keyframe — more keyframes do not automatically mean more control. Every extra one is another point the model has to hit exactly, and the motion between closely spaced ones has nowhere to go.

The four ways it goes wrong

Kling lists the common mistakes, and each one has a screenwriting equivalent. Too many keyframes: over-blocking a scene until no one can move. Changing too many things at once between two keyframes — costume, light, location and action in one step — which a cut could handle but a continuous shot cannot. Inconsistent references: keyframes and reference images that disagree, so a character looks different from one moment to the next. And conflicting prompts, where the words pull one way and the pictures another, and the creative direction, as Kling puts it, becomes unclear.

A way to start

Write the scene as you would for a reader first. Mark the moments where the situation changes — those are your keyframes, and a 30-second scene rarely has more than five. Draw or generate one image for each. Then write the timeline: what moves between each pair, in present tense, with the camera's part named. Kling 4.0 is in limited early access with the full model due in October, so a plan written this way now is ready the day the account has it.