🖨️ MoneyPrinter V2 — Enhanced Edition

AI-powered content creation platform. Generate tech videos, YouTube Shorts, and Twitter posts — fully self-hosted, no paid LLM APIs required.

Features · Architecture · Setup · Configuration · Roadmap

---

🙏 Built on top of the excellent MoneyPrinterV2 by @FujiwaraChoki. The original project provided the complete automation foundation — YouTube Shorts generation, Twitter posting, affiliate marketing, and local business outreach. Without that base, none of this would have been possible. Go give it a ⭐.

---

🆕 What's New

This fork keeps everything from the original CLI and adds a full web platform on top:

Layer	What was added
🎨 React Dashboard	Visual job management, real-time pipeline view, video player
⚙️ FastAPI Backend	REST API + WebSocket, PostgreSQL, Celery job queue
🎬 Remotion Renderer	Node.js microservice that renders animated MP4s with React
🤖 Tech Video Pipeline	9-step AI pipeline: search → script → images → TTS → subtitles → render
📹 YouTube Shorts Pipeline	10-step pipeline with web search, duration control, topic override, and auto-upload
🔗 Account Connect Flow	One-click Firefox session capture — log in once, all uploads automated
💰 Cost Tracking	Per-service cost breakdown (fal.ai, Ollama, edge-tts, Whisper)

---

🎬 Tech Video Pipeline

The flagship new feature. Triggered from the dashboard, the full pipeline runs automatically:

🔍 Web Search          →  Tavily API (optional — fetch current docs/news)
📝 Generate Script     →  Ollama (Qwen2.5-Coder recommended)
🏷️  Generate Metadata  →  YouTube title + description
💻 Code Snippets       →  2–3 real examples (tech-dark template)
🖼️  Image Prompts      →  LLM-generated prompts
🎨 Generate Images     →  fal.ai FLUX (~$0.003/image)
🔊 Synthesize Voice    →  edge-tts (7 languages, Microsoft Neural)
📋 Word Subtitles      →  faster-whisper (TikTok-style, per-word timing)
🎥 Render Video        →  Remotion 4 + Chromium → MP4

Each step streams real-time progress to the dashboard via WebSocket.

Templates

Template	Style	Best for
tech-dark	Dark gradient + animated code blocks + Ken Burns images	Programming tutorials
minimal	Full-bleed images + vignette overlay	Clean explainers
bold	Title card intro + flash cuts + accent bar	High-energy content
reel	Animated gradient backgrounds + pulse highlight	Short-form reels

Resolutions

Key	Size	Platform
shorts	1080 × 1920	YouTube Shorts · TikTok · Reels
landscape	1920 × 1080	YouTube standard
square	1080 × 1080	Instagram

---

📹 YouTube Shorts Pipeline

✅ Fully working end-to-end — tested on Ubuntu 24.04 with dual-GPU (NVIDIA + AMD). See Troubleshooting if Firefox doesn't open.

Generate, compose, and upload in one click:

🔍 Web Search          →  Tavily API (optional — improves script accuracy)
💡 Generate Topic      →  Ollama LLM (skipped if topic provided manually)
📝 Generate Script     →  Ollama LLM (~75–300 words depending on duration)
🏷️  Generate Metadata  →  YouTube title + description
🖼️  Image Prompts      →  LLM-generated prompts (3–6 images)
🎨 Generate Images     →  fal.ai FLUX schnell (~$0.003/image)
🔊 Synthesize Voice    →  edge-tts (7 languages, Microsoft Neural)
📋 Generate Subtitles  →  faster-whisper base (word-level SRT)
🎞️  Compose Video      →  MoviePy — images + TTS + background music + subtitles
📤 Upload to YouTube   →  Selenium Firefox — auto or manual re-upload from dashboard

Option	Values	Default
Duration	30s · 60s · 90s · 120s	60s
Language	English, Spanish, Portuguese, French, German, Italian, Japanese	English
Topic	Free text (skips topic generation step)	—
Web search	on / off	off
Auto-upload	on / off	off

If auto-upload is off (or the upload step fails), a "Upload to YouTube" button appears on the job card — click it to upload the already-rendered video without regenerating anything.

Account Connect Flow

No credentials stored. Session captured once from a plain (non-automated) Firefox:

1. Create an account in the dashboard (YouTube or Twitter)
2. Click Connect → plain Firefox opens to the platform login page
3. Log in normally — Google/YouTube does not block this (no WebDriver involved)
4. Dashboard reads the session cookie from the Firefox profile and saves the path
5. All future jobs reuse that profile — fully automated

Real-time status streams back via WebSocket: opening → waiting → detected → connected.

---

🏗️ Architecture

┌─────────────────────────────────────────────────────┐
│                   React Dashboard                   │  :5173
│  Dashboard · YouTube · Twitter · Tech Videos        │
│  Costs · Config · Real-time pipeline via WebSocket  │
└────────────────────┬────────────────────────────────┘
                     │ HTTP / WebSocket
┌────────────────────▼────────────────────────────────┐
│              FastAPI Backend                        │  :8001
│  Jobs · Videos · Accounts · Costs · Config · WS    │
│  PostgreSQL + Redis Pub/Sub                         │
└──────┬──────────────────────────┬───────────────────┘
       │ Celery tasks             │ HTTP POST
┌──────▼──────────┐    ┌──────────▼──────────────────┐
│  Celery Workers │    │  Remotion Service (Node.js)  │  :3001
│  • youtube      │    │  Express · Webpack bundle    │
│  • twitter      │    │  4 templates · 3 resolutions │
│  • remotion     │    │  Renders MP4 via Chromium    │
└─────────────────┘    └─────────────────────────────┘
         │
┌────────▼────────────────────────────────────────────┐
│                  Local Services                     │
│  🦙 Ollama  ·  🐘 PostgreSQL  ·  🔴 Redis           │
│  🎨 fal.ai  ·  🔊 edge-tts  ·  👂 faster-whisper   │
└─────────────────────────────────────────────────────┘

---

🧰 Stack

Layer	Technology
Frontend	React 18, Vite, Tailwind CSS, shadcn/ui
Backend	FastAPI, SQLAlchemy (async), PostgreSQL, Alembic
Task Queue	Celery + Redis
Real-time	Redis Pub/Sub → WebSocket
LLM	Ollama — Qwen2.5-Coder, Llama3, Mistral, etc.
Image Gen	fal.ai (FLUX)
TTS	edge-tts (Microsoft Neural voices, 7 languages)
STT	faster-whisper (word-level timestamps)
Video Render	Remotion 4 — React → MP4 via headless Chromium
Web Search	Tavily API (optional)
Browser Automation	Selenium + pre-authenticated Firefox profiles

💡 The only paid external services are fal.ai (~$0.003/image) and Tavily (free tier available). Everything else runs locally.

---

📋 Requirements

• 🐍 Python 3.12
• 🟢 Node.js 18+
• 🐘 PostgreSQL
• 🔴 Redis
• 🦙 Ollama with at least one model pulled
• 🐹 Go — only needed for Local Business Outreach

---

🚀 Setup

1. Clone

git clone https://github.com/luisfarfan/MoneyPrinterV2.git
cd MoneyPrinterV2

2. Python environment

cp config.example.json config.json   # then fill in your values
python -m venv venv
source venv/bin/activate             # Windows: .\venv\Scripts\activate
pip install -r requirements.txt

3. Remotion microservice

cd remotion-service
npm install
cd ..

4. React dashboard

cd frontend
npm install
cd ..

5. Environment variables

cp .env.example .env

Edit .env:

DATABASE_URL=postgresql+asyncpg://mpv2:mpv2@localhost:5432/mpv2
REDIS_URL=redis://localhost:6379/0
FAL_KEY=your_fal_api_key
TAVILY_API_KEY=your_tavily_api_key   # optional

6. Database

alembic upgrade head

7. Pull an LLM model

ollama pull qwen2.5-coder

---

▶️ Running

Open 4 terminals from the project root:

# Terminal 1 — FastAPI backend
uvicorn backend.main:app --host 0.0.0.0 --port 8001

# Terminal 2 — Celery worker
celery -A backend.workers.celery_app worker --loglevel=info -Q remotion,youtube,twitter -c 1

# Terminal 3 — Remotion renderer
cd remotion-service && npx tsx src/server.ts

# Terminal 4 — React dashboard
cd frontend && npm run dev

Open http://localhost:5173 🎉

macOS quick setup

bash scripts/setup_local.sh
python scripts/preflight_local.py

Docker (partial)

docker-compose up

---

⚙️ Configuration

config.json (Python CLI)

Key	Description
ollama_model	LLM model name (e.g. qwen2.5-coder:latest)
imagemagick_path	Required for MoviePy subtitle rendering
firefox_profile	Path to pre-authenticated Firefox profile

.env (Backend + Remotion)

Key	Description
DATABASE_URL	PostgreSQL connection string
REDIS_URL	Redis connection string
FAL_KEY	fal.ai API key — image generation
TAVILY_API_KEY	Tavily API key — web search (optional)

---

🗺️ Roadmap

✅ Done

• Web search integration for YouTube Shorts pipeline (Tavily)
• Duration selector per job (30s / 60s / 90s / 120s)
• Topic override (skip LLM topic generation)
• Account connect flow — plain Firefox session capture (no WebDriver, Google won't block)
• Auto-upload YouTube Shorts via Selenium after generation
• Manual re-upload button for failed/skipped upload steps
• Real-time pipeline progress via WebSocket
• Cost tracking per service and per job
• Firefox dual-GPU fix for Ubuntu 24.04 (NVIDIA + AMD)
• Twitter auto-post from dashboard — generate tweet (Ollama) + post via Selenium in one click

🚧 Next Up

• 🎵 TikTok — connect account + auto-upload Shorts (same Selenium flow as YouTube)
• 📤 Auto-upload Remotion tech videos to YouTube from dashboard

📅 Planned

• 📤 Auto-upload Remotion tech videos to YouTube
• ⏰ CRON scheduling for automated posting
• 🖼️ Video thumbnail previews in job list
• 💸 Cost budget alerts and per-account limits
• 🐳 Full Docker Compose setup (one command for everything)
• 📱 Mobile-responsive dashboard
• 🔑 Support OpenAI / Anthropic as LLM providers alongside Ollama
• 🖥️ Virtual display (Xvfb) for headless server deployments

---

📁 Project Structure

MoneyPrinterV2/
├── src/                        # Original Python CLI (unchanged)
│   ├── main.py                 # Interactive menu
│   ├── cron.py                 # Headless scheduler runner
│   └── classes/                # YouTube, Twitter, AFM, Outreach, TTS
│
├── backend/                    # FastAPI backend ✨
│   ├── models/                 # SQLAlchemy ORM (Job, Video, Account, Cost)
│   ├── routes/                 # REST API routers
│   ├── schemas/                # Pydantic request/response schemas
│   ├── workers/                # Celery tasks (youtube, twitter, remotion)
│   └── settings.py
│
├── remotion-service/           # Node.js video renderer ✨
│   └── src/
│       ├── server.ts           # Express API (/render, /health, /media)
│       ├── renderer.ts         # Remotion bundle + renderMedia logic
│       ├── types.ts            # Zod schemas + shared types
│       ├── compositions/       # TechDark, Minimal, Bold, Reel
│       └── components/         # GradientBackground, SubtitleWord, ImageScene…
│
├── frontend/                   # React dashboard ✨
│   └── src/
│       ├── pages/              # Dashboard, YouTube, Twitter, TechVideo, Costs, Config
│       ├── components/         # VideoPreview, PipelineView, AccountRow
│       ├── hooks/              # useJobStream (WebSocket)
│       └── lib/                # api.ts — typed fetch client
│
├── docs/
├── scripts/
├── config.example.json
└── .env.example

---

🛠️ Troubleshooting

Common issues and fixes are documented in docs/Troubleshooting.md.

Issue	Quick link
Firefox won't open / Process unexpectedly closed with status 0	Firefox crash on Ubuntu 24.04 dual-GPU
Firefox profile path not found	Profile path error

---

🙏 Acknowledgments

Project	Why
MoneyPrinterV2 by @FujiwaraChoki	The entire original automation foundation this fork is built on
Remotion	React-based video rendering engine
Ollama	Local LLM inference
fal.ai	Fast image generation API
edge-tts	Microsoft Neural TTS voices
faster-whisper	Word-level speech transcription

---

📄 License

Licensed under the GNU Affero General Public License v3.0 — same as the original project. See LICENSE for details.

⚠️ Disclaimer

For educational purposes only. All automation must comply with the terms of service of the platforms being accessed. The authors are not responsible for any misuse.