TU

themattberman/ugc-factory-skill

API development
16 stars Quality 55 Trend 21MIT

using Kling 3.0 (via fal.ai). Generate authentic-feeling AI character videos—testimonials, unboxings, demos—through a 4-module pipeline with approval gates. No real creators required.

Overview

using Kling 3.0 (via fal.ai). Generate authentic-feeling AI character videos—testimonials, unboxings, demos—through a 4-module pipeline with approval gates. No real creators required. - → Brief (format, hook, character, setting) - → Visual beat sheet (shot-by-shot, no dialogue yet) - → Script synced to shot durations - → fal.ai JSON + execution → video Visuals before script. Design how each shot looks and moves first, then write dialogue that fits. The script serves the visuals. Kling 3.0’s system locks character identity across videos, so you can build a consistent “creator” library. - — Kling 3.0 video generation. Get an API key at fal.ai. - An image model (e.g. for character reference images). The skill references “Nano Banana Pro” via an /ai-image-generation skill if you have one; you can also supply your own reference image URLs. This repo contains no API keys or secrets. You supply a fal.ai key via the FAL_KEY environment variable only. .env is gitignored; use .env.

README

UGC Content Factory

Director-first UGC content generation using Kling 3.0 (via fal.ai). Generate authentic-feeling AI character videos—testimonials, unboxings, demos—through a 4-module pipeline with approval gates. No real creators required.

By Matthew Berman — @themattberman · Big Players · Emerald Digital


What It Does

  • Creative Director → Brief (format, hook, character, setting)
  • Cinematographer → Visual beat sheet (shot-by-shot, no dialogue yet)
  • Screenwriter → Script synced to shot durations
  • Engineer → fal.ai JSON + execution → video

Principle: Visuals before script. Design how each shot looks and moves first, then write dialogue that fits. The script serves the visuals.

Kling 3.0’s Elements system locks character identity across videos, so you can build a consistent “creator” library.


Requirements

  • fal.ai — Kling 3.0 video generation. Get an API key at fal.ai.
  • Optional: An image model (e.g. for character reference images). The skill references “Nano Banana Pro” via an /ai-image-generation skill if you have one; you can also supply your own reference image URLs.

Credentials: This repo contains no API keys or secrets. You supply a fal.ai key via the FAL_KEY environment variable only. .env is gitignored; use .env.example as a template. Do not commit keys or .env.


Installation

1. Clone the repo

git clone https://github.com/themattberman/ugc-content-factory.git
cd ugc-content-factory

2. Install the skill (Cursor / Claude Code)

Copy the skill folder into your AI assistant’s skills directory so the skill is available by name (e.g. “ugc content factory” or “ugc-content-factory”):

  • Cursor: e.g. ~/.cursor/skills/ or your project’s skills path — copy the entire Skill/ugc-content-factory folder (including references/ and SKILL.md).
  • Claude Code / Codex: e.g. ~/.claude/skills/ — same: copy Skill/ugc-content-factory so that SKILL.md and references/ sit inside one ugc-content-factory (or ugc_content_factory) folder.

Exact paths depend on your setup; the important part is that the skill’s SKILL.md and references/ directory are together and recognized by your tool.

3. Set your API key

# Option A: export in shell (e.g. add to ~/.zshrc or ~/.bashrc)
export FAL_KEY="your-fal-ai-api-key"

# Option B: copy .env.example to .env, add your FAL_KEY, then load .env in your shell
cp .env.example .env

Get your key from fal.ai (dashboard or API settings).

4. (Optional) Test fal.ai

From the repo root, with FAL_KEY set:

cd Skill/ugc-content-factory
./test-fal-api.sh

Requires curl and jq. The script runs a simple text-to-video call to confirm your key and endpoint work.


Usage

Invoke the skill with phrases like:

  • ugc / ugc video / ugc creator
  • testimonial video / user generated content

The skill will:

  1. Ask whether you have existing assets (character image URL, product images, seed frame).
  2. Run Module A (Creative Director) and present a Creative Brief → Gate 1 (you approve).
  3. Run Module B (Cinematographer) and present a Visual Beat Sheet → Gate 2 (you approve).
  4. Run Module C (Screenwriter) and present a Synchronized Script → Gate 3 (you approve).
  5. Run Module D (Engineer) to build fal.ai payloads and execute generation.

Reference docs live in Skill/ugc-content-factory/references/ (e.g. MODULE_A_CREATIVE_DIRECTOR.md through MODULE_D_ENGINEER.md, plus CHARACTER_LIBRARY.md and SCENE_SETTINGS.md). The skill instructs the AI to read these when running each phase.


Repo Layout

ugc-content-factory/
├── README.md
├── LICENSE
├── .gitignore
├── .env.example
└── Skill/
    └── ugc-content-factory/
        ├── SKILL.md              # Main skill definition and pipeline
        ├── test-fal-api.sh       # Optional: test fal.ai connection
        ├── TESTING-PLAN.md       # Testing checklist for the pipeline
        └── references/
            ├── MODULE_A_CREATIVE_DIRECTOR.md
            ├── MODULE_B_CINEMATOGRAPHER.md
            ├── MODULE_C_SCREENWRITER.md
            ├── MODULE_D_ENGINEER.md
            ├── CHARACTER_LIBRARY.md
            └── SCENE_SETTINGS.md


License

MIT. See LICENSE.

View the full README on GitHub

Recommended Tools

Try a different keyword or remove a filter.

Install

npx skillfish add themattberman/ugc-factory-skill
New to MCP? Start with the mcpskill.org guide →