An AI-powered system that automatically generates animated educational videos from text prompts using Google ADK agents, Manim animations, and ElevenLabs narration.

Status: Private Repository | Proprietary Project

• System Architecture
• Prerequisites
• Installation
• Configuration
• Running the Agent
• How It Works
• Project Structure
• Troubleshooting
• Roadmap
• Support
• Contributing
• License

The pipeline consists of three main agent groups working in sequence:

• Manages the entire pipeline flow
• Coordinates between transcript and video generation
• Handles state management across all sub-agents
• Will manage watermark agent in future updates

Topic Researcher → Content Structurer → Script Writer → Speech Formatter 
    ↓
Audio Transcriber ← ElevenLabs TTS
    ↓
Scene Summary

Video Agent → [Manim Code Gen, Render Checker, Scene Validator]
    ↓
Concatenator (combines all scenes + audio + music)
    ↓
[Future: Watermark Agent → Final branded video]

• Python 3.8+ installed and in PATH
• FFmpeg for video/audio processing
• Manim Community Edition for animations
• API Keys for:
  • Google ADK/Gemini
  • ElevenLabs (TTS)
  • Groq (STT)

git clone <your-repo-url>
cd educational-video-pipeline

# Windows
python -m venv venv
venv\Scripts\activate

# Mac/Linux
python3 -m venv venv
source venv/bin/activate

pip install -r requirements.txt

• Windows: Download from ffmpeg.org, extract, and add to PATH
• Mac: brew install ffmpeg
• Linux: sudo apt install ffmpeg (Ubuntu/Debian) or sudo yum install ffmpeg (Fedora)

# Should be installed via requirements.txt, but if issues arise:
pip install manim

# Check FFmpeg
ffmpeg -version

# Check Manim
manim --version

# Check Python packages
python -c "import elevenlabs, groq, manim; print('All packages installed!')"

Create a .env file in the project root:

# API Keys
ELEVENLABS_API_KEY=your_elevenlabs_key_here
GROQ_API_KEY=your_groq_key_here
GOOGLE_API_KEY=your_google_key_here  # If using Google services

Edit join.py and update the Manim executable path:

# Line ~41 in join.py
MANIM_EXECUTABLE = r"C:\path\to\your\manim.exe"  # Windows
# or
MANIM_EXECUTABLE = "/usr/local/bin/manim"  # Mac/Linux

Place your background music file at:

C:\Users\leg\Documents\elearn\agent\music\background.mp3

Or update the path in the concatenate agent instruction.

python join.py

When prompted, paste your educational video request. Format:

Create a [duration]-minute [level] educational video about [topic].
Include [specific requirements].

Example:
Create a 3-minute beginner-level educational video about photosynthesis.
Include colorful animations and simple analogies for middle school students.

The pipeline will:

1. Research the topic (30-60 seconds)
2. Generate script and narration (1-2 minutes)
3. Create and render animations (2-5 minutes per scene)
4. Concatenate final video (30 seconds)

Total time: 5-15 minutes depending on video length

The completed video will be in:

final_videos/educational_video_[timestamp].mp4

1. Topic Researcher - Searches for comprehensive information
2. Content Structurer - Creates timed educational outline
3. Script Writer - Converts outline to conversational script
4. Speech Formatter - Adds pauses and timing for TTS
5. ElevenLabs TTS - Generates professional narration
6. Audio Transcriber - Creates timestamped transcription
7. Scene Summary - Breaks content into animated scenes

1. Video Agent coordinates the creation of each scene:
   • Manim Code Gen - Generates animation code
   • Render Checker - Executes and monitors rendering
   • Scene Validator - Ensures timing matches narration

1. Concatenator - Combines all scenes with audio and background music
2. [Future] Watermark Agent - Adds branding and outro

educational-video-pipeline/
├── join.py              # Main application orchestrator
├── prompt.py            # Agent instructions and prompts
├── stt_tools.py         # Speech-to-text functionality
├── tts_tool.py          # Text-to-speech functionality
├── .env                 # API keys (create from .env.example)
├── requirements.txt     # Python dependencies
├── generated_audio/     # TTS output files
├── transcriptions/      # STT transcriptions
├── media/              # Manim render files
├── final_videos/       # Completed videos
└── music/
    └── background.mp3   # Background music

• Verify Manim installation: manim --version
• Update MANIM_EXECUTABLE path in join.py
• Try full path: C:\Python39\Scripts\manim.exe

• Check .env file has valid API keys
• Wait 30 seconds and retry
• The system has automatic retry logic (3 attempts)

• Ensure FFmpeg is in system PATH
• Test with: ffmpeg -version
• Restart terminal after PATH updates

• System automatically adjusts timing ±0.5 seconds
• For larger mismatches, scenes are regenerated
• Maximum 30 attempts per scene

# Reinstall all dependencies
pip install --upgrade -r requirements.txt

# For specific package issues
pip uninstall [package_name]
pip install [package_name]

To see detailed logs, run with:

python join.py > debug.log 2>&1

Edit tts_tool.py to change voices:

voice_id = "21m00Tcm4TlvDq8ikWAM"  # Rachel (default)
# Other options:
# "EXAVITQu4vr4xnSDxMaL" - Bella
# "ErXwobaYiN019PkySvjV" - Antoni

Modify prompts in prompt.py to change animation preferences

• Volume: Adjust music_volume in concatenate agent (default: 0.15)
• File: Update path in concatenate agent instruction

• OpenAI TTS Integration - Add support for OpenAI's text-to-speech models
• Google Cloud TTS - Integrate Google's WaveNet voices
• Amazon Polly - Add AWS Polly for more voice variety
• Voice Cloning - Support for custom voice cloning with ElevenLabs
• Multi-language Support - Enable video generation in multiple languages

• Watermark Agent - New automated branding agent

  Concatenator → Watermark Agent → Final Output
  

  Features:

  • Auto-detect video dimensions and add appropriate watermark
  • Support for image (PNG/SVG) and video watermarks
  • Configurable position (corner/center/custom)
  • Fade in/out animations
  • Duration control (full video or last X seconds)

  Example workflow:

  # The agent will automatically:
  1. Take the concatenated video
  2. Apply watermark based on config
  3. Add outro video if specified
  4. Output final branded video

• Outro Integration

  • Pre-made outro videos with smooth transitions
  • Dynamic text overlay (channel name, social links)
  • Subscribe button animations
  • End screen templates matching video style

• Quality Presets - Platform-specific optimization
  • YouTube (1080p/4K horizontal)
  • TikTok/Shorts (9:16 vertical)
  • Instagram Reels (9:16 with safe zones)
  • Twitter/X video specifications
• Interactive Elements - Clickable areas in videos
• Multiple Video Formats - Support for vertical videos (Shorts/Reels)
• Batch Processing - Generate multiple videos from CSV input
• Cloud Deployment - Deploy pipeline to cloud services
• Web Interface - Browser-based UI for easier access

🔨 Working on: Watermark Agent implementation for automatic branding

For issues or questions:

1. Check the Troubleshooting section
2. Review error messages in the console
3. Ensure all prerequisites are properly installed

This is a private repository. Contributions are limited to authorized team members only.

This is a private repository. All rights reserved. No part of this software may be reproduced, distributed, or transmitted without prior written permission.

Built with ❤️ using Google ADK, Manim, ElevenLabs, and Groq