
themattberman/ugc-factory-skill
API developmentこのページは AI により自動翻訳されています。技術的な詳細は原文をご確認ください。 原文を見る
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.
概要
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
以下は原文 README の抜粋です。原文を見る。
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-generationskill 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 entireSkill/ugc-content-factoryfolder (includingreferences/andSKILL.md). - Claude Code / Codex: e.g.
~/.claude/skills/— same: copySkill/ugc-content-factoryso thatSKILL.mdandreferences/sit inside oneugc-content-factory(orugc_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:
推奨ツール
別のキーワードを試すか、フィルタを外してください。
インストール
npx skillfish add themattberman/ugc-factory-skill