YouTube Creator Automation Pipeline

An automated Python pipeline designed to generate engaging, high-fidelity short-form videos for YouTube Shorts and Facebook Reels.

The pipeline leverages Google Cloud Vertex AI (Gemini, Veo 3.1, Imagen 3.0) and Cloud Text-to-Speech to generate all media assets headlessly. It uses local FFmpeg for assembly and includes browser automation via Playwright for Facebook Reels uploads.

Features

• Brain: Gemini 2.5 Flash generates unique daily scripts and media prompts.
• Video Hook: Veo 3.1 generates a cinematic 8-second 9:16 vertical video hook.
• Image Fill: Imagen 3.0 generates high-quality vertical visuals to complete the timeline.
• Narration: Google Cloud TTS (Journey voice) for human-like narration.
• Subtitles: Local OpenAI Whisper model for accurate SRT generation.
• Assembly: Dynamic FFmpeg engine with Ken Burns effects, crossfades, and ambient background music.
• Multi-Platform: Automated publishing to YouTube and Facebook Reels.

Prerequisites

• Python 3.10+
• FFmpeg installed in system PATH.
• Google Cloud Project with Vertex AI and Cloud TTS enabled.

---

Detailed Configuration Guide

1. Google Cloud / Vertex AI / YouTube

To use the automated generation and YouTube uploading features, you must configure a Google Cloud Project:

1. Create a GCP Project: Go to the Google Cloud Console and create a new project.
2. Enable APIs: In the "APIs & Services" dashboard, enable the following:
   • Vertex AI API (Required for Gemini, Veo, and Imagen)
   • Cloud Text-to-Speech API (Required for narration)
   • YouTube Data API v3 (Required for uploading)
3. Application Default Credentials (ADC): Install the Google Cloud CLI and run:

   gcloud auth application-default login

   This allows main.py to authenticate with Vertex AI and TTS services.
4. YouTube OAuth Credentials:
   • Go to APIs & Services > Credentials.
   • Click Create Credentials > OAuth client ID.
   • Select Desktop App as the application type.
   • Download the JSON file and rename it to client_secrets.json in the project root.
   • Initial Run: Execute python youtube_uploader.py. This will open a browser for a one-time manual authorization. It generates token.json, which the pipeline uses for headless uploads thereafter.

2. Facebook Reels (Meta)

The pipeline uses Playwright browser automation to upload Reels, as the official Meta Graph API for Reels is highly restricted.

1. Initial Authentication:
   • The first time the pipeline runs (or if you run python meta_uploader.py directly), Playwright will launch a browser.
   • If no session is found, it will pause and prompt you in the terminal to log in to Facebook manually in the opened browser window.
   • After logging in and reaching the home page, press Enter in your terminal.
2. Session Persistence:
   • The pipeline saves your login state to fb_state.json.
   • Subsequent runs will use this file to upload headlessly without requiring manual intervention.
   • Note: If fb_state.json expires or the upload fails due to a login wall, simply delete fb_state.json and run the pipeline again to re-authenticate.

3. Audio Library

• Add your .mp3 or .wav background tracks to the audio_library/ directory.
• The pipeline selects one at random for each video and mixes it with the narration at a lower volume.

---

Installation

# Clone the repository
git clone <repository-url>
cd Youtube_Creator_Automation

# Install dependencies (recommended to use a venv)
python -m venv venv
source venv/bin/activate
pip install .
playwright install chromium

Running the Pipeline

Manual Execution

export GOOGLE_CLOUD_PROJECT="your-project-id"
python main.py

Automated Execution (Linux/Systemd)

Use the provided deployment script to schedule the pipeline daily at 3:00 AM:

./deploy_systemd.sh your-project-id

Project Structure

• main.py: The central orchestrator.
• youtube_uploader.py: YouTube Data API integration & OAuth manager.
• meta_uploader.py: Facebook Reels browser automation via Playwright.
• ledger.py: SQLite-based history to avoid duplicate topics.
• audio_library/: Directory for background music assets.
• output/: Temporary directory for generated media.

License

MIT