ai-video-production

name	ai-video-production
description	Turns a validated creative brief into scene-by-scene prompts for AI video generation tools (VEO3, Kling, Runway). Use when producing video with AI tools instead of human editors. Trigger on: 'VEO3 prompts', 'Kling prompt', 'generate AI video', 'AI video brief'. Produces a scene-by-scene prompt pack ready to execute in the chosen AI tool.
metadata	{"version":"1.0.0","status":"beta","tags":["production","ai-video","VEO3","Kling","prompts"],"inputs":["validated creative brief","selected hook","reference images if available"],"outputs":["scene-by-scene prompt pack for VEO3 / Kling"],"depends-on":["creative-brief","hook-writing"]}

AI Video Production

Before Starting

Confirm before starting:

• Validated creative brief
• Selected hook
• AI tool defined: VEO3 (Google) or Kling (Kuaishou) or Runway
• Target duration per scene defined (recommended: 3–8s per scene)
• Reference images available if visual consistency is required (character consistency)
• Visual style defined (realistic / cinematic / simulated UGC / animated)

---

AI Video Prompting Principles

Scene Prompt Structure

[Subject] [Action] [Setting] [Camera angle] [Camera movement] [Lighting] [Style] [Mood / Emotion] [Duration]

Rules:

• 1 scene = 1 prompt. Do not combine two actions in the same prompt.
• Name the action before the context: "A woman opens [object]" not "In a kitchen, a woman..."
• Specify camera movement: static / slow zoom / tracking / handheld / drone
• Specify lighting: natural / golden hour / studio / low key / overcast
• Avoid abstract words: "beautiful", "amazing" — describe what is concrete

Visual Consistency Across Scenes

To keep the same character across scenes:

• Physically describe the character in each prompt (age, gender, outfit, hair color)
• Use a reference image if the tool allows it (Kling = image + prompt)
• Keep the same setting / background across linked scenes

---

Phase 1: Break the Brief into Scenes

Take the narrative structure from the brief (Hook → Tension → Resolution → CTA) and convert it into 3–8 second scenes.

Recommended number of scenes:

• 15s → 3–4 scenes
• 30s → 5–6 scenes
• 60s → 8–10 scenes

---

Phase 2: Write the Prompts

For each scene:

**Scene [#] — [Name / Role in structure]**
Duration: [3–8s]
Tool: [VEO3 / Kling / Runway]

Prompt:
"[Complete description following the structure: subject + action + setting + camera + lighting + style + emotion]"

Reference image: [Yes / No — link if yes]
Production note: [Specific advice for this tool]

---

Phase 3: Assemble and Test

After generating all scenes:

1. Watch each scene individually
2. Check visual consistency (character, setting, style)
3. Assemble in narrative order
4. Add text overlay, subtitles, music in post (CapCut / Premiere / DaVinci)
5. Check: does the sequence hold without sound? (silent viewing test)

---

Output Format

# AI Video Prompt Pack — [Brand] — [Creative ID] — [Date]

## Context
Primary tool: [VEO3 / Kling / Runway]
Total duration: [Xs]
Style: [Realistic UGC / Cinematic / Animated]
Main character: [Complete physical description]

## Scenes

### Scene 1 — Hook [0–3s]
Prompt: "[Full prompt]"
Ref image: [Link / None]

### Scene 2 — [Name] [3–Xs]
[Same structure]

[...]

## Post-Production
- Text overlay to add: "[Text hook]" — [Position, font, timing]
- Subtitles: [Yes / No]
- Music: [Style / Reference]
- Final CTA: "[Text]" — [Appearance timing]

---

Related Skills

• 03-strategy/creative-brief — prerequisite
• 03-strategy/hook-writing — hook → first scene
• 04-production/video-production — human editor brief version of the same concept
• 04-production/campaign-setup — after production