Video Production

Video Production Skill

Text-based video production expertise for YouTube and social video platforms. All outputs are text — no actual video/image generation.

Core Methodology: Hook-First

Every video script follows the hook-first pattern:

1. Hook (0-5s): Pattern interrupt that stops the scroll. Must create curiosity gap or emotional trigger.
2. Setup (5-30s): Context + promise of value. State exactly what the viewer will learn/get.
3. Body (30s-end): Deliver value in clear segments with retention triggers every 60-90 seconds.
4. CTA (final 15s): Single clear action. Never stack multiple CTAs.

Retention Triggers

Insert one every 60-90 seconds to prevent drop-off:

• Pattern interrupt: Change visual pace, topic shift, "but here's where it gets interesting"
• Open loop: Tease upcoming content: "I'll show you the exact template in a moment"
• Social proof: Quick data point or result
• Direct address: "Now you might be thinking..." — acknowledge viewer's mental state
• B-roll cue: [CUT TO: screen recording / example / diagram]

Script Writing Format

markdownCopy code
# [VIDEO TITLE — YouTube optimized, <60 chars]

**Target length:** [X] minutes
**Audience:** [specific viewer persona]
**Core promise:** [single sentence — what viewer gets]

---

## HOOK (0:00 - 0:05)

[TALKING HEAD]
"[Pattern interrupt — question, bold statement, or surprising fact]"

## SETUP (0:05 - 0:30)

[TALKING HEAD + B-ROLL]
"[Context. What this video covers. Why it matters NOW.]"

**Retention checkpoint:** [open loop for upcoming content]

## SECTION 1: [Topic] (0:30 - X:XX)

[SCREEN RECORDING / B-ROLL]
"[Teaching content with specific, actionable steps]"

**Key point:** [One-sentence takeaway]

[Repeat sections as needed]

## CTA (X:XX - end)

[TALKING HEAD]
"[Single action: subscribe, link in description, comment]"

---

### SEO Metadata
- **Title:** [<60 chars, front-load keyword]
- **Description:** [see description template below]
- **Tags:** [10-15 relevant tags]
- **Chapters:** [timestamp markers from sections]

YouTube Description Template

markdownCopy code
[Hook sentence — repeat the core promise from the video]

In this video, I [action verb — show/teach/reveal/break down] [specific topic].

🔗 Resources mentioned:
- [Resource 1]: [URL]
- [Resource 2]: [URL]

⏱️ Chapters:
0:00 — [Hook/Intro]
0:30 — [Section 1 title]
X:XX — [Section 2 title]
[...]

📌 Key takeaways:
• [Takeaway 1]
• [Takeaway 2]
• [Takeaway 3]

[CTA paragraph — what to do next, related videos]

#[tag1] #[tag2] #[tag3]

Thumbnail Concept Spec

Thumbnails are described as text specs for sharp/SVG generation. Never attempt to generate actual images.

yamlCopy code
thumbnail:
  layout: [split | centered | thirds | before-after]
  dimensions: "1280x720"
  background:
    type: [solid | gradient | image-blur]
    value: "[color hex or gradient spec]"
  text:
    headline: "[3-5 words MAX — large, readable at mobile size]"
    font_weight: bold
    font_size: "~120px equivalent"
    color: "[high contrast against background]"
    position: [left | center | right]
    stroke: "[optional outline color for readability]"
  face:
    expression: "[emotion: surprise, excitement, curiosity, shock]"
    position: [left | right]
    size: "[head + shoulders, ~40% of frame]"
  accent_elements:
    - type: [arrow | circle | emoji | icon]
      position: "[x, y quadrant]"
      color: "[brand accent color]"
  contrast_score: "[high | medium — must be high for mobile feeds]"
  text_readability: "[mobile-first — readable at 100px thumbnail width]"

Thumbnail Best Practices

• 3-word rule: Headline text should be 3-5 words maximum
• Face + emotion: Human faces with exaggerated expressions increase CTR 30-40%
• High contrast: Must be readable as a 100px-wide thumbnail on mobile
• No clickbait: Thumbnail must honestly represent video content
• Brand consistency: Use recurring color palette across video series
• A/B variants: Always produce 2 thumbnail concepts for testing

Video SEO

Tags Strategy

yamlCopy code
tags:
  primary: "[exact target keyword]"
  secondary:
    - "[keyword variation 1]"
    - "[keyword variation 2]"
    - "[related topic 1]"
  long_tail:
    - "[question-format keyword]"
    - "[how-to format keyword]"
  brand:
    - "[channel name]"
    - "[series name if applicable]"
  total: 10-15  # sweet spot for YouTube

Chapter Markers

• First chapter MUST start at 0:00
• Minimum 3 chapters for YouTube to display them
• Each chapter title: concise, keyword-rich, <40 chars
• Chapters appear in Google search results — optimize for search intent

YouTube Algorithm Signals

Storyboarding

yamlCopy code
storyboard:
  scene_number: [N]
  timestamp: "[start - end]"
  visual:
    type: [talking-head | screen-recording | b-roll | animation | text-overlay]
    description: "[what viewer sees]"
  audio:
    narration: "[script text for this segment]"
    music: "[mood: upbeat, chill, dramatic, none]"
    sfx: "[sound effect if any]"
  text_overlay: "[on-screen text, lower-thirds, callouts]"
  transition: "[cut | fade | zoom | swipe]"
  retention_device: "[hook | open-loop | proof | direct-address | none]"
  notes: "[production notes, emphasis points]"

Output Types

Anti-Patterns

• Starting with "Hey guys, welcome back" — hook first, intro later
• Stacking multiple CTAs — one CTA per video
• Ignoring mobile thumbnail readability
• Generic tags like "tutorial" or "how to"
• Description with only links, no searchable text
• Chapters without 0:00 start marker
• Scripts without retention triggers in long-form content

Integration with Remotion Rendering

This skill produces text-based storyboards and scripts. To convert a storyboard into an actual MP4 video, hand off to the remotion-video skill and the video-producer agent, which handle:

1. Storyboard → Composition mapping (remotion-video skill)
2. Remotion environment discovery (auto-detects CLI, entry point, and deps on any machine)
3. Render to MP4 (video-producer agent Phase 3)

Handoff flow:

Copy code
video-production skill (this skill) → generates storyboard YAML
  ↓
remotion-video skill → maps to composition JSON (frame calculations, animations)
  ↓
video-producer agent → discovers Remotion CLI, renders MP4, opens video

The storyboard YAML format from this skill is compatible with the remotion-video composition mapping when these fields are present per scene: type, duration (seconds), text, animation, transition. The remotion-video skill calculates startFrame and durationFrames automatically.