Y
Prompt Workshop
Example · Video shot

AI Video Shot Prompt Teardown: 5-Second Rain-Night Still

This teardown is the safest video prompt to start with: static camera + single micro motion + explicit stability. Five seconds, clean physics, very high success rate across Runway, Pika, Kling, Seedance and Hailuo.

The full prompt

a single raindrop slides slowly down a foggy window at night, camera holds completely static, neon city lights blurred in the background, slow motion 120fps look, single vertical motion only, shallow depth of field, cool blue and amber color grading, no camera shake, 5-second cinematic clip

Structure breakdown

actiona single raindrop slides slowly down — one vertical motion
cameracamera holds completely static — no drift
environmentfoggy window + neon city lights blurred
speedslow motion 120fps look
motion locksingle vertical motion only — no horizontal sway
depthshallow depth of field
gradecool blue and amber
stabilityno camera shake + 5-second cinematic clip

The reason this prompt works is the combination: static camera + single slow micro motion. AI video models hit 70%+ success here. Add multi-axis camera motion or multiple actions and the success rate often drops below 20%.

3 rewrites

Rewrite 1 · Tracking shotCamera moves with subject
a young woman in a black trench coat walks slowly forward through a rain-soaked Tokyo alley at night, camera tracks horizontally to the right at the same walking pace, neon reflections on wet ground, shallow depth of field, steady gimbal feel, 6-second cinematic shot, no jitter

Tracking shot: subject walks + camera follows. "At the same walking pace" stops the camera from overtaking the subject.

Rewrite 2 · Drone pushLandscape, no people
misty mountain valley at sunrise, slow steady drone dolly forward over the treetops, sunlight breaking through low clouds, very gentle pacing over 8 seconds, cinematic wide shot, smooth motion, no jitter

Drone push: "slow steady drone dolly forward" + "over 8 seconds" controls the push speed.

Rewrite 3 · Food macroSingle dripping action
extreme close-up of melted chocolate slowly dripping onto a glossy croissant, camera holds completely static, single dripping motion only, warm side light from the right, shallow depth of field, slow motion 120fps look, 3-second clip, no shake

Food macros work as static + single slow action. "Single dripping motion only" blocks stirs and other parasitic motions.

Common pitfalls

Pitfall 1 · Vague action verbs

"Moves", "interacts" are too soft. Use "slides slowly", "pours", "turns to the left".

Pitfall 2 · Multiple actions in one shot

"Walks in, sits down, picks up" in 5 seconds will fail. One core action per shot.

Pitfall 3 · No duration

Without "3-second / 5-second clip", the model invents a default pacing that is usually too fast.

Pitfall 4 · No stability

Without "no camera shake / steady", most models add jitter that breaks the still-life mood.

Pitfall 5 · Subject and camera both moving fast

Heavy subject motion + heavy camera motion collapses the shot.

Pitfall 6 · Image quality keywords reused

"Masterpiece, 8k, best quality" do almost nothing for video. "Cinematic, shallow depth of field, color grading" is enough.

Model comparison

ModelTypical durationStrengthRating on this shot
Runway Gen-35–10 sCinematic, tracking★★★★★ static-shot king
Pika 2.x3–5 sShort atmospheric★★★★ great in 3–5 s
Kling 2.x5–10 sPeople, ads★★★★ excellent non-English support
Seedance 2.05–8 sWidescreen cinematic★★★★★ best for 16:9
Hailuo / MiniMax5–6 sPeople, landscapes★★★ long descriptive prompts

Where this skeleton fits

Use this skeleton for: opening atmosphere shots, mid-piece product ad b-roll, social-media story headers, trailer transitions. The general rule: static camera + single slow action is the highest-success AI video shape. For multi-shot stories, edit several stills together rather than asking one prompt to deliver a complex sequence.

Frequently asked questions

Why do AI videos sometimes look jelly-like or skip frames?

Almost always because the action is too complex or the duration is too long. Drop to 3–5 seconds, one action, static camera and the success rate jumps.

Do video prompts need negative prompts?

Most models barely use them. Use positive constraints like "no camera shake", "single action only".

How do I keep a character consistent across multiple video shots?

Drive each shot from the same reference image (image-to-video) and lock the seed where supported.

Can I write video prompts in languages other than English?

Kling and Hailuo handle Chinese extremely well; Runway, Pika and Sora prefer English.

Save this prompt to your library

The library supports stars, tags, notes and TXT / Markdown / CSV export — all in your browser.

Open the library →
Yan · AI Prompt Workshop editorial team|Last updated on 2026-06-12。This site does not call any cloud model. Every prompt and parameter in this article was tested and refined locally by the editorial team.