ugc-content-factory

📁 themattberman/ugc-factory-skill 📅 5 days ago
4
总安装量
4
周安装量
#48970
全站排名
安装命令
npx skills add https://github.com/themattberman/ugc-factory-skill --skill ugc-content-factory

Agent 安装分布

opencode 4
gemini-cli 4
github-copilot 4
codex 4
kimi-cli 4
amp 4

Skill 文档

UGC Content Factory

Generate authentic-feeling UGC content at scale with consistent AI characters.

Use this skill when: You need UGC-style video content (testimonials, reviews, unboxings, demos) without hiring real creators.

The unlock: Kling 3.0’s Elements system (via fal.ai) locks character identity across multiple videos. Combined with multi-shot storyboards and native lip-sync, you can generate an entire UGC library with consistent “creators.”


Director-First Pipeline

Key principle: Visuals before script. Design how each shot looks and moves FIRST, then write dialogue that fits. The script serves the visuals, not the other way around.

MODULE A: Creative Director    → Creative Brief        → GATE 1 (user approves)
                                                          │
MODULE B: Cinematographer      → Visual Beat Sheet      → GATE 2 (user approves)
                                                          │
MODULE C: Screenwriter         → Synchronized Script    → GATE 3 (user approves)
                                                          │
MODULE D: Engineer             → fal.ai JSON + Execute  → Output

Variable durations: Each shot can be 3s, 4s, or 5s. Hooks are punchy (3s), demos breathe (5s). Total must sum to target duration.


Before You Start: Existing Assets Check

Before entering the pipeline, ASK the user:

  1. “Do you have an existing character to reuse? (Provide Element image URL)”
  2. “Do you have product images to use as Elements? (Provide image URL)”
  3. “Do you have a seed frame / first-frame image? (Provide image URL)”

If yes → use those URLs in the pipeline. If no → new images will be generated with Nano Banana Pro via /ai-image-generation.

Check references/CHARACTER_LIBRARY.md for any previously saved characters/products.


How to Execute

Phase A: Creative Director

Read references/MODULE_A_CREATIVE_DIRECTOR.md before proceeding.

Also load for reference:

  • references/CHARACTER_LIBRARY.md (character archetypes + saved characters)
  • references/SCENE_SETTINGS.md (setting options)

Tasks:

  1. Select content format (Testimonial, Unboxing, Problem/Solution, Comparison, Demo)
  2. Choose hook angle (Gatekeep, Skeptic, Fail, Visual Shock)
  3. Select character archetype and setting (or reuse saved character)
  4. Run Stealth Validation (4 mandatory checks)
  5. Present Creative Brief to user

⛔ GATE 1: Present Creative Brief. STOP. Wait for user approval before proceeding.


Phase B: Cinematographer

Read references/MODULE_B_CINEMATOGRAPHER.md before proceeding.

Tasks:

  1. Design shot-by-shot visual sequence (NO dialogue)
  2. Assign variable durations (3s/4s/5s per shot)
  3. Apply Kling prompt engineering (6 elements, 4 rules)
  4. Verify scene variety (2-of-4 position/angle changes)
  5. Generate seed images if using I2V workflow (Nano Banana Pro via /ai-image-generation — use Photorealistic Pre-Prompt)
  6. Present Visual Beat Sheet to user

⛔ GATE 2: Present Visual Beat Sheet. STOP. Wait for user approval before proceeding.


Phase C: Screenwriter

Read references/MODULE_C_SCREENWRITER.md before proceeding.

Tasks:

  1. Calculate word budget per shot (~2.5 words/sec)
  2. Write dialogue fitted to shot durations
  3. Apply 4 Audio Rules (sync, cadence, room sound, bridge)
  4. Map script to Beat Sheet (script-to-beat-sheet template)
  5. Run script quality check
  6. Present Synchronized Script to user

⛔ GATE 3: Present Synchronized Script. STOP. Wait for user approval before proceeding.


Phase D: Engineer

Read references/MODULE_D_ENGINEER.md before proceeding.

Tasks:

  1. Assemble Visual Beat Sheet + Script into multi_prompt JSON
  2. Run compilation checklist
  3. Submit to fal.ai (standard for test, pro for final)
  4. Use queue endpoint for anything over 5 seconds
  5. Present output with quality check

Character Setup (One-Time)

Before first video with a new character:

  1. Define persona brief (see CHARACTER_LIBRARY.md)
  2. Generate 3-5 reference images via Nano Banana Pro (/ai-image-generation) — use Photorealistic Pre-Prompt from MODULE_B
  3. Select best reference image(s)
  4. Upload to fal.ai Kling 3.0 Elements
  5. Save character to CHARACTER_LIBRARY.md “Saved Characters” section for reuse

Integration with Other Skills

Skill Use For
/ai-image-generation Character reference images via Nano Banana
/ai-talking-head Fallback for single-shot videos
/hooks Generate hook variations for A/B testing
/direct-response-copy Script optimization for conversion
/brand-voice Ensure scripts match client voice

The 80/20

80% of UGC success comes from:

  1. Right character for target audience
  2. Specific, genuine-sounding script (not marketing speak)
  3. Appropriate casual setting (not studio)
  4. Strong hook in first 2 seconds

Get these right and the rest is optimization.


Quick Tips

Do:

  • Lead with imperfection — slightly off framing, natural pauses, filler words
  • Use specific details — “3 weeks” not “a while”
  • Match character to audience demographic
  • Keep it short — 15-30 seconds, not 60+

Don’t:

  • Over-polish — if it looks like an ad, it fails
  • Use marketing language — “revolutionary” = instant scroll-past
  • Skip the visual design phase — visuals drive everything
  • Fight the shot durations — adapt the script to the visuals

Module Reference Files

All detailed instructions are in references/:

File Content Lines
MODULE_A_CREATIVE_DIRECTOR.md Formats, hooks, stealth validation, creative brief ~250
MODULE_B_CINEMATOGRAPHER.md Visual design, Kling prompting, beat sheet ~300
MODULE_C_SCREENWRITER.md Audio rules, sync, script templates, mapping ~250
MODULE_D_ENGINEER.md API reference, JSON assembly, queue management ~200
CHARACTER_LIBRARY.md Archetypes, design process ~100
SCENE_SETTINGS.md Setting tables and selection rules ~50