Content Engine

Automated short-form video content pipeline for micro-SaaS distribution.

Generate TikTok/Reels/Shorts videos programmatically — script generation, text-to-speech, video composition, all automated. Ship a product, plug in a config, generate content at scale. Kill what doesn't work, keep what does.

Topic → Claude Code (script) → Edge TTS (voice) → Remotion (video) → MP4

---

Philosophy

The pipeline is the business. Products are experiments.

• Ship fast: 1-2 weekends per product MVP
• Kill fast: 30 days to prove or kill
• Content is distribution: 5-10 videos/day across multiple accounts
• $0 marginal cost: Claude Code Max + Edge TTS + Remotion = free per video

---

Quick Start

# Clone & install
git clone https://github.com/AirKyzzZ/content-engine.git
cd content-engine
npm install

# Create your first product
npm run new-product -- my-saas

# Edit the config
# → src/products/my-saas/config.ts

# Generate a video
npm run generate -- my-saas "your topic here"

# Batch generate from default topics
npm run batch -- my-saas

# Download gameplay backgrounds
npm run download -- my-saas

---

Architecture

content-engine/
├── src/
│   ├── engine/              # Core engine (product-agnostic)
│   │   ├── scripts.ts       # Claude Code CLI → structured scripts
│   │   ├── voice.ts         # Edge TTS + ffmpeg post-processing
│   │   ├── gameplay.ts      # YouTube download + 9:16 crop
│   │   └── render.ts        # Remotion bundle + render
│   │
│   ├── compositions/        # Video templates (React components)
│   │   ├── ChatOverBroll    # Tinder-style chat over gameplay
│   │   ├── TextOverBroll    # Text callouts over gameplay
│   │   └── shared/          # Reusable: ChatBubble, PhoneChat, etc.
│   │
│   ├── products/            # Product configs (one folder each)
│   │   └── rizzia/config.ts # Example: AI dating coach
│   │
│   └── cli.ts               # CLI entry point
│
├── public/                  # Assets (gameplay, voice, music)
└── out/{product-id}/        # Rendered MP4s per product

---

Product Config

Each product is a single config.ts file:

const config: ProductConfig = {
  id: "my-saas",
  brandName: "MySaaS",
  handle: "@mysaas",
  description: "What does this product do",
  audience: "Who is the target (age, interests, pain points)",
  formats: ["chat-over-broll"],
  ctaPlacement: "mid", // mid-roll CTA at sec 8-12 (3-5x better than end)

  promptInstructions: `Master prompt for script generation...`,

  topics: [
    "Default topic 1",
    "Default topic 2",
  ],

  gameplayClips: [
    { url: "https://youtube.com/watch?v=...", tag: "subway-surfer" },
  ],

  voices: {
    male: "fr-FR-HenriNeural",
    female: "fr-FR-DeniseNeural",
    narrator: "fr-FR-HenriNeural",
    rate: "+15%",
  },
};

---

CLI Commands

Command	Description
npm run generate -- <product> "topic"	Generate one video
npm run batch -- <product> [topics...]	Batch generate (uses default topics if none provided)
npm run download -- <product>	Download gameplay clips from YouTube
npm run new-product -- <id>	Scaffold a new product config
npm run studio	Open Remotion Studio for visual preview

---

Stack

Component	Tool	Cost
Script generation	Claude Code CLI (claude -p)	$0 (Max subscription)
Text-to-speech	Edge TTS (Microsoft)	$0 (free, no API key)
Audio processing	ffmpeg (EQ + compression + normalize)	$0
Video composition	Remotion (React-based)	$0 (individuals)
Gameplay footage	yt-dlp (YouTube CC0 clips)	$0
Posting	Postiz (self-hosted)	$0

Total cost per video: $0

---

Video Features

• Phone-style chat UI (dark mode, avatar, status bar, message input)
• Male + female voices reading messages (synced to visual)
• Full-screen gameplay with subtle drift animation
• Screen shake on message impacts
• Configurable mid-roll or end-roll CTA
• Emoji reactions (displayed but not spoken by TTS)
• "@handle" watermark
• Audio post-processing (EQ, compression, -14 LUFS normalization)

---

Requirements

• Node.js 20+
• ffmpeg + ffprobe
• yt-dlp (for gameplay downloads)
• edge-tts (Python CLI: pipx install edge-tts)
• Claude Code CLI (for script generation)

---

Roadmap

• ProfileScore composition (AI rates your profile)
• BeforeAfter composition (profile transformation)
• Carousel/slideshow composition (static tip cards)
• Background music layer
• Kokoro TTS integration (better French quality)
• Analytics collector (YouTube API + TikTok API)
• Feedback loop (LLM analyzes top/bottom performers)
• Multi-account posting via Postiz

---

License

MIT