Imported from rymqan/staple-tech (.claude/skills/video-prompting/SKILL.md). Install upstream with npx skills add rymqan/staple-tech --skill video-prompting. Copyright stays with the author.

Video Prompting for Social Media Marketing

This skill encodes production-grade video prompt engineering knowledge specifically for short-form social media marketing videos (5-15 seconds). The ScriptWriterAgent's system prompt references this skill to produce prompts that video generation APIs interpret correctly.

Section 1 — The 6-Part Prompt Formula

Every video prompt follows this structure in order:

Subject → Action → Environment → Camera → Style → Constraints

Subject

Describe WHO or WHAT is in the frame — visually, not conceptually.

• Good: "woman in black leather jacket with silver rings, mid-30s, dark wavy hair"
• Bad: "a woman" or "our customer"
• Good: "artisan sourdough loaf with golden crust and flour dusting"
• Bad: "bread" or "our product"

Action

One present-tense verb per shot. The model renders exactly what you describe.

• Good: "turns slowly toward camera"
• Bad: "dynamic movement" or "engages with the product"
• Good: "pours steaming espresso into a ceramic cup"
• Bad: "makes coffee"

Environment

WHERE the action happens + atmospheric detail the model can render.

• Good: "sunlit industrial kitchen with exposed brick walls and copper pans hanging overhead"
• Bad: "in a kitchen" or "nice background"

Camera

Shot size + one movement + angle. Never stack two movements.

• Good: "medium close-up, slow dolly-in, eye level"
• Bad: "dynamic camera work" or "cinematic angles"

Style

One visual anchor: a film stock, lighting setup, or color grade.

• Good: "warm 35mm film grain, golden hour backlight"
• Bad: "cinematic" or "professional looking"

Constraints

Ban list, duration, aspect ratio, consistency notes.

• Always include: "Avoid jitter, bent limbs, morphing faces, abrupt cuts, inconsistent lighting"
• Specify: duration (5s/10s/15s), aspect ratio (9:16/1:1/16:9)

Target: 60-100 words per shot segment.

Section 2 — 2-Second Hook Framework

The first 2 seconds determine if a viewer keeps watching. Choose ONE hook type per video:

1. Scale Shift

Giant or miniature reveal — an object at impossible scale. Prompt: "Extreme close-up of [object] filling the entire frame. Slow pull-back reveals it is [context showing unexpected scale]."

2. Impossible Physics

Object defying gravity, then snapping to reality. Prompt: "Static tripod. [Object] suspended mid-air, slowly rotating. Beat. Falls naturally into [surface], landing with impact."

3. Dramatic Light Reveal

Darkness to light cutting through. Prompt: "Black frame. Single beam of warm light sweeps left to right, gradually illuminating [subject] from shadow."

4. Extreme Close-Up Pull-Back

Macro detail expanding to wide context. Prompt: "Extreme close-up of [texture/detail], filling frame completely. Smooth dolly-out over 2 seconds to reveal full [subject] in [environment]."

5. Freeze-to-Motion

Still frame that comes alive. Prompt: "Static frame: [subject] frozen mid-[action], completely still. Beat of 0.5 seconds. Subject comes to life, completing the [action] smoothly."

6. Product Materialization

Object assembling from particles or fragments. Prompt: "Empty frame, [background]. [Product particles/fragments] swirl inward from edges, assembling into [complete product] at center frame."

7. Before/After Split

Split-screen transformation. Prompt: "Split frame: left side shows [before state], right side shows [after state]. Divider sweeps left to right, transforming the entire frame to [after state]."

8. Emotional Reaction Close-Up

Face reacting to off-screen reveal. Prompt: "Close-up of [person description], neutral expression, looking slightly off-camera. Eyes widen subtly, corners of mouth lift into genuine smile."

9. Time-Lapse Collapse

Hours compressed to 2 seconds. Prompt: "High angle, static tripod. [Process] plays at 60x speed: [describe the visual transformation from start to end in 2 seconds]."

10. Perspective Impossible

Camera passes through solid object. Prompt: "Camera pushes forward smoothly, passing through [solid surface — glass, wall, door] to reveal [new environment on other side]."

Section 3 — Timeline Segmentation

Format: [0s-3s], [3s-7s], [7s-12s], [12s-15s]

Rules

• 3-5 segments for a 15-second video
• 2-3 segments for a 10-second video
• 1-2 segments for a 5-second video
• One camera movement per segment
• One action verb per segment
• Global setup (subject, style, lighting, negative constraints) goes BEFORE the timeline, not inside it

Transitions Between Segments

• cut — hard cut, most common
• crossfade — soft blend, for mood transitions
• whip pan — fast horizontal sweep, for energy
• match cut — visual element carries between shots

Template Structure

[Global: subject description, style anchor, lighting setup]
[Negative constraints: avoid jitter, bent limbs, morphing faces...]
[Duration: 15s | Aspect: 9:16]

[0s-3s]: [shot size], [camera movement]. [Subject] [action]. [Environment detail].
[3s-7s]: [shot size], [camera movement]. [Subject] [action]. [Environment detail].
[7s-12s]: [shot size], [camera movement]. [Subject] [action]. [Environment detail].
[12s-15s]: [shot size], [camera movement]. [Subject] [action]. [Environment detail].

Section 4 — Camera Movement Vocabulary

Use ONLY these terms — video generation models are trained on this vocabulary.

Movement

• dolly in / dolly out — camera physically moves forward/backward
• push in / pull back — same as dolly but implies faster pace
• pan left / pan right — camera rotates horizontally on fixed point
• tilt up / tilt down — camera rotates vertically on fixed point
• tracking shot — camera follows subject laterally
• crane up / crane down — camera rises/lowers vertically
• orbit / arc — camera circles the subject
• whip pan — fast horizontal rotation (transition move)
• static tripod — camera does not move (stabilized)
• handheld — adds subtle micro-wobble (UGC, documentary feel)
• gimbal — smooth stabilized movement (cinematic feel)

Shot Size

• extreme close-up (ECU) — fills frame with detail (eye, texture, small object)
• close-up (CU) — face or single object fills frame
• medium close-up (MCU) — head and shoulders
• medium (MS) — waist up
• medium wide (MWS) — knees up
• wide (WS) — full body with environment
• extreme wide (EWS) — landscape, subject is small in frame

Angle

• eye level — neutral, standard perspective
• low angle — camera below subject, looking up (power, authority)
• high angle — camera above subject, looking down (vulnerability, overview)
• bird's eye — directly overhead
• worm's eye — directly below, looking straight up
• dutch angle — camera tilted off-axis (tension, unease)

Lens Feel

• wide lens (24-28mm) — expands space, slight distortion at edges
• normal lens (35-50mm) — natural perspective, versatile
• telephoto (85mm+) — compressed depth, isolates subject, creamy bokeh

Compound Moves

Express as sequential beats, never simultaneous:

• "Start: slow dolly-in. Then: gentle pan right for final 2 seconds."
• "Begin: static tripod. At 3s: crane up to reveal skyline."

CRITICAL RULE

One motion verb per shot. Two simultaneous camera verbs = visual chaos. If you need compound motion, sequence them as beats.

Section 5 — Speed and Quality Rules

"Fast" is the single word most likely to degrade video generation quality.

Rules

1. Never combine fast camera + fast cuts + busy scene in the same segment
2. Default to slow, smooth, gentle for all camera movement
3. Only ONE element should be fast — if the action is fast, the camera should be slow
4. Speed words need distance/scale: "slow dolly-in, 1-2 feet" not just "slow dolly-in"
5. If pace is needed, achieve it through editing (shorter segments, cuts) not camera speed

Speed Hierarchy (safest to riskiest)

1. static tripod — safest, zero motion artifacts
2. slow [movement] — reliable across all models
3. smooth [movement] — reliable, implies steady pace
4. gentle [movement] — reliable, subtle motion
5. steady [movement] — moderate pace, generally safe
6. [movement] (no modifier) — model decides speed, unpredictable
7. fast [movement] — high artifact risk, use sparingly

Section 6 — Lighting (Highest-Leverage Single Addition)

One lighting description improves output quality more than ten adjectives. Always include lighting.

8 Marketing-Relevant Lighting Setups

1. Soft Ring Light

Use for: UGC, beauty, testimonials, skincare Prompt phrase: "Even soft ring light illuminating the face, catchlights in both eyes, minimal shadows"

2. Golden Hour Backlight

Use for: Lifestyle, outdoor, aspirational, wellness Prompt phrase: "Warm golden hour backlight from behind, lens flare at frame edge, long soft shadows"

3. High-Key Flat Lighting

Use for: Product shots, clean/modern, tech, minimalist Prompt phrase: "Bright high-key flat lighting, even illumination, no harsh shadows, white or neutral background"

4. Dramatic Side Light (Rembrandt)

Use for: Premium, luxury, editorial, fashion Prompt phrase: "Single strong side light from the left, deep shadows on the right, dramatic Rembrandt lighting"

5. Neon Rim Light

Use for: Tech, gaming, youth culture, nightlife Prompt phrase: "Neon rim lighting in [color] from behind, dark environment, subject silhouette edged with color"

6. Warm Ambient / Candlelight

Use for: Food, cozy, intimate, restaurant, home Prompt phrase: "Warm ambient candlelight glow, soft flickering, deep warm tones, intimate atmosphere"

7. Overcast Diffused

Use for: Fashion, street style, casual, urban Prompt phrase: "Soft overcast diffused daylight, no harsh shadows, even natural illumination, muted tones"

8. Studio Three-Point

Use for: Professional, corporate, interview, product demo Prompt phrase: "Classic three-point studio lighting: key light from 45 degrees, fill light opposite, backlight rim"

Section 7 — Negative Constraints (Standard Defaults)

Always include for character-based videos:

Avoid: jitter, bent limbs, morphing faces, extra fingers, abrupt cuts,
inconsistent lighting between segments, text overlays, watermarks,
unnatural skin texture, flickering shadows.

For product-focused videos, add:

Avoid: product deformation, floating objects, inconsistent product color,
blurred logos, unrealistic reflections.

For environment/landscape, add:

Avoid: warping buildings, flickering sky, inconsistent horizon line,
repeated textures, swimming clouds.

Section 8 — Platform Format Rules

Instagram Reels / TikTok

• Aspect ratio: 9:16
• Duration: 15 seconds max (generate 15s clips)
• Hook: first 2 seconds must stop the scroll
• Audio: native Seedance audio or uploaded voiceover
• Safe zone: keep subject in center 80% (UI overlays on edges)

Instagram Feed Video

• Aspect ratio: 1:1
• Duration: 15 seconds max
• Hook: first 2 seconds
• Autoplay: muted by default — visual hook matters more than audio

YouTube Shorts

• Aspect ratio: 9:16
• Duration: 60 seconds max (but generate 15s clips)
• Hook: first 2 seconds
• Similar to Reels but slightly more tolerant of slower pacing

Instagram Stories

• Aspect ratio: 9:16
• Duration: 15 seconds per story frame
• Full-screen vertical — use entire frame
• Tap-through UX — front-load the message

LinkedIn Video

• Aspect ratio: 16:9
• Duration: 15-30 seconds
• Professional tone — slower pacing, cleaner compositions
• Text-heavy environment — strong visual contrast needed

Section 9 — 5 Production-Ready Example Prompts

Example 1: Product Showcase (E-Commerce, Unboxing Reveal)

Global: Matte-finish luxury candle in amber glass jar with gold embossed label.
Style: Warm 35mm film grain, soft shallow depth of field, desaturated warm palette.
Lighting: Warm ambient candlelight glow, soft flickering, deep warm tones.
Negative: Avoid jitter, morphing product shape, inconsistent label text,
flickering shadows, text overlays.
Duration: 15s | Aspect: 9:16

[0s-3s] HOOK — Product Materialization:
Extreme close-up, static tripod. Black frame. Golden particles swirl inward
from edges, assembling into the candle jar at center frame. Warm light blooms
from the wick as the candle materializes completely.

[3s-7s] DETAIL:
Close-up, slow dolly-in. Camera approaches the gold embossed label on the amber
glass. Candlelight flickers across the raised lettering. Shallow depth of field
blurs the background to warm bokeh.

[7s-12s] ENVIRONMENT:
Medium wide, gentle orbit left. Candle sits on a dark walnut bedside table.
Linen sheets and a leather-bound book visible in soft focus. Evening light
through sheer curtains. Warm intimate bedroom atmosphere.

[12s-15s] HERO:
Medium close-up, static tripod. Flame dances gently. Wisps of scented smoke
curl upward. Warm amber glow fills the frame. Camera holds steady on the
beauty shot.

Example 2: UGC Testimonial (Person Speaking to Camera)

Global: Young woman, early 20s, natural makeup, oversized cream knit sweater,
holding a skincare bottle with green label. Genuine, unscripted energy.
Style: iPhone-quality handheld footage, natural color, slight overexposure.
Lighting: Soft ring light illuminating the face, catchlights in both eyes,
minimal shadows. Bedroom background slightly overexposed.
Negative: Avoid morphing face, bent fingers, unnatural mouth movement,
jitter, inconsistent skin texture, text overlays.
Duration: 15s | Aspect: 9:16

[0s-3s] HOOK — Emotional Reaction:
Close-up, handheld. Woman looks directly at camera, eyes wide with excitement.
Holds up the skincare bottle beside her face. Subtle head shake of disbelief.
Genuine smile breaks across her face.

[3s-7s] SHOW:
Medium close-up, handheld with micro-wobble. Woman tilts the bottle to show
the label to camera. Points at it with her other hand. Natural gestures,
relaxed posture. Ring light catchlights in her eyes.

[7s-12s] APPLY:
Close-up, static tripod. Woman's hands squeeze product onto fingertips.
Dabs gently onto her cheek. Smooth blending motion. Soft natural skin texture.
Product has visible creamy consistency.

[12s-15s] REACTION:
Medium close-up, handheld. Woman touches her cheek, looks at camera.
Nods approvingly. Holds the bottle up one final time beside her face.
Natural bathroom lighting mixes with ring light.

Example 3: Food/Beverage Hero (Preparation, Plating, Pour)

Global: Artisan sourdough bread with deep golden crust, flour dusting, and open crumb.
Professional food photography aesthetic.
Style: Clean editorial food photography, warm color grade, shallow depth of field.
Lighting: Side light from the left at 45 degrees, creating texture shadows on
the crust. Fill light from the right, soft and diffused.
Negative: Avoid morphing bread shape, unrealistic steam, flickering crumb texture,
inconsistent crust color, text overlays, watermarks.
Duration: 15s | Aspect: 9:16

[0s-3s] HOOK — Extreme Close-Up Pull-Back:
Extreme close-up of the sourdough crust texture, filling the entire frame.
Every bubble and crack visible. Slow dolly-out over 2 seconds reveals the
full loaf sitting on a wooden cutting board.

[3s-7s] THE CUT:
Medium close-up, static tripod. Hands enter frame from below holding a
serrated bread knife. Slow deliberate sawing motion through the crust.
The bread yields, revealing the open honeycomb crumb structure inside.
Flour particles drift in the side light.

[7s-12s] REVEAL:
Close-up, slow dolly-in. The cut cross-section fills the frame. Open airy
crumb structure with irregular holes. Steam rises gently from the warm
interior. Side light catches the moisture on the crumb walls.

[12s-15s] SERVE:
Medium, gentle pan right. A hand places a thick slice onto a ceramic plate.
Butter knife spreads golden butter across the warm surface. Butter melts
on contact. Rustic kitchen background in soft focus.

Example 4: Brand Story Origin (Founder, Workspace, Craft)

Global: Male artisan woodworker, 40s, leather apron, calloused hands, focused expression.
Small woodworking studio with sawdust and hand tools.
Style: Documentary 16mm film feel, muted earth tones, natural grain.
Lighting: Overcast diffused daylight from large workshop windows. No harsh shadows.
Supplemented by a single warm work lamp on the bench.
Negative: Avoid morphing face, bent fingers on tools, jitter, warping wood grain,
inconsistent sawdust, text overlays.
Duration: 15s | Aspect: 9:16

[0s-3s] HOOK — Dramatic Light Reveal:
Black frame. A warm beam of light sweeps across from left, gradually illuminating
the woodworker's hands resting on a raw plank of walnut. Sawdust particles
float in the light beam. Workshop emerges from darkness.

[3s-7s] CRAFT:
Close-up, slow tracking shot right. Hands guide a hand plane along the walnut
board. Thin wood shaving curls upward in a continuous ribbon. Steady controlled
motion. Work lamp casts warm directional light across the grain.

[7s-12s] FOCUS:
Medium close-up, static tripod. The artisan examines the smoothed surface,
running fingertips along the grain. Tilts the board to catch the light.
Satisfied nod. Workshop shelves with finished pieces visible in soft background.

[12s-15s] RESULT:
Medium wide, slow crane up. The finished piece — a minimalist walnut serving
board — sits on the workbench. Tools arranged neatly beside it. The artisan
steps back, removes his leather gloves. Workshop in full view.

Example 5: Social Hook/Trend (Scroll-Stopping Opener)

Global: Minimalist white desk setup with a sleek laptop, ceramic coffee mug,
and a single green plant. Clean modern workspace.
Style: High-key flat lighting, crisp digital clarity, cool-neutral color grade.
Lighting: Bright high-key flat lighting, even illumination, no harsh shadows.
Supplemented by cool blue monitor glow on the laptop screen.
Negative: Avoid warping desk surface, flickering screen, jitter, morphing plant,
inconsistent mug position, text overlays.
Duration: 15s | Aspect: 9:16

[0s-3s] HOOK — Freeze-to-Motion:
Medium, static tripod. Everything frozen: coffee mid-pour into the mug,
laptop screen glowing, plant leaf mid-sway. Complete stillness for 1 second.
Then everything comes to life — coffee splashes into the mug, leaf completes
its sway, cursor moves on screen.

[3s-7s] ROUTINE:
Close-up, smooth dolly-in. Fingers type rapidly on the laptop keyboard.
Screen reflects in the user's glasses (no face visible, just the reflection).
Coffee steam rises from the mug in the foreground, soft focus.

[7s-12s] REVEAL:
Medium, gentle orbit right. The full desk setup comes into view: laptop
centered, wireless mouse to the right, notebook open to the left with
handwritten notes. Plant in the corner. Clean cable management.
Everything aligned and intentional.

[12s-15s] MOMENT:
Close-up, static tripod. Hand wraps around the ceramic mug. Lifts it
slowly to just out of frame. Steam trail follows the motion upward.
Laptop screen glows steadily in the background. Quiet productive energy.