Verified against Kling 3.0 · 2026-08-10
Generate a two-second hook clip built to stop a thumb mid-scroll
A Kling 3.0 prompt focused entirely on the opening beat of a video — the fraction of a second that decides whether a viewer keeps scrolling — rather than treating the hook as the first few seconds of a longer script.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Open cold, mid-action, no establishing shot: a stack of plates wobbling mid-fall before being caught is already happening the instant the clip starts, as if the viewer opened the app one second too late to see it begin. Within the first beat, introduce the hand catching it belongs to someone walking by, not the person who dropped it so the eye has something to resolve immediately rather than settle. The camera does one thing only — a quick whip-pan from the falling stack to the catching hand — fast enough to feel urgent but not so fast the subject smears. Set this in a busy restaurant kitchen pass, lit so the main subject is unmistakably the brightest thing in frame even in a small phone thumbnail. No slow build, no scene-setting pan, no title card: the first frame and the last frame of this clip should each work as a standalone still that makes someone want to know what happens next.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Feed algorithms and viewer attention both punish anything that reads as a wind-up, so instructing Kling 3.0 to start mid-action instead of establishing the scene removes the single most common reason generated hooks underperform: a beautiful but slow first two seconds that a thumb has already scrolled past by the time anything interesting happens. Naming exactly one camera move, and pairing it with an explicit note about not smearing the subject, works with rather than against Kling's known motion trade-off — fast whip-pans are one of the areas where video models are most prone to losing subject shape, so calling it out directly reduces the chance the model either slows the move down (killing the urgency) or blurs the payoff (killing the reveal). The instruction that both the first and last frame should work as a standalone still is a practical proxy for thumbnail and loop performance — it forces genuinely strong, legible composition at both boundary frames rather than a clip that only looks good in motion, which is what actually gets tested when a platform auto-selects a cover frame.
What you get back
A tight 2-second clip: a stack of plates already mid-tip over a kitchen pass, whip-pan reveals a passing server's hand catching them a half-second before they'd have shattered, bright pass-light keeping the catch unmistakably legible even at thumbnail size.
Verified against
Kling 3.0 3.0 · 2026-08-10
Changelog
- 2026-08-10 — Initial publish, verified against Kling 3.0.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
