A web application that generates mathematical animations using Manim (Mathematical Animation Engine) based on user prompts, powered by Hugging Face's Mixtral 8x7B model.
- AI-Powered Animation Generation: Utilizes Hugging Face's Mixtral 8x7B model to convert natural language prompts into Manim code
- Real-Time Rendering Pipeline: Asynchronous video processing with Flask backend
- Code Sanitization Engine: Robust code cleaning and validation system
- Fallback Mechanism: Graceful degradation with default animations when generation fails
- Task Management System: Background job tracking with UUID-based task identification
- Python 3 with Flask web framework
- Manim (Mathematical Animation Engine) for vector graphics rendering
- Hugging Face Inference API (Mixtral 8x7B model)
- Subprocess Management for Manim CLI execution
- Thread-based Asynchronous Processing
- UUID-based Task Tracking
- HTML5 semantic markup
- CSS3 with responsive design principles
- JavaScript ES6+ with Fetch API for asynchronous operations
- AJAX-based Status Polling
- File System Management for temporary storage and video output
- Environment Variable Configuration for API keys
- Cross-Platform Path Handling
- Python 3.8+
- Manim Community Edition (
pip install manim) - FFmpeg
- Hugging Face API key (set as
HF_API_KEYenvironment variable)
# Clone the repository
git clone https://github.com/yourusername/manim-generator.git
cd manim-generator
# Create and activate virtual environment
python -m venv venv
source venv/bin/activate # Linux/Mac
venv\Scripts\activate # Windows
# Install dependencies
pip install -r requirements.txt# Set your Hugging Face API key
export HF_API_KEY="your_api_key_here" # Linux/Mac
set HF_API_KEY="your_api_key_here" # WindowsGradientlearning/
|__Authentication
|__app.py
| # Flask application entry point
├── static/
│ ├── videos/ # Rendered animation storage
│ └── css/ # Static CSS files
├── templates/
│ └── index.html # Frontend interface
└── temp_scenes/ # Temporary Manim script storage
-
AI Integration Layer
- Utilizes Hugging Face's inference API with Mixtral 8x7B model
- Implements prompt engineering with specific constraints for Manim code generation
- Includes retry mechanism (3 attempts) with exponential backoff
-
Code Processing Engine
- AST-based code sanitization
- Syntax validation and automatic indentation correction
- Class name collision avoidance system
- Fallback code generation subsystem
-
Rendering Pipeline
- Subprocess-based Manim execution
- Media file management with automatic cleanup
- Quality level configuration (currently set to low quality
-ql)
-
Asynchronous Task Management
- Thread-based parallel processing
- UUID-based task tracking
- Status polling endpoint with JSON responses
| Endpoint | Method | Description |
|---|---|---|
/ |
GET | Serves the frontend interface |
/render |
POST | Initiates animation generation (accepts JSON with prompt) |
/check-result |
GET | Checks task status (requires task_id parameter) |
python app.pyThe application will be available at http://localhost:5000
- Video rendering is CPU-intensive - consider queue systems for production
- Temporary files are automatically managed but not auto-purged
- Low-quality rendering (
-qlflag) balances speed and quality
- API keys are loaded from environment variables
- Input sanitization is performed on generated code
- UUIDs prevent predictable resource access
- Complex prompts may generate invalid Manim code
- Rendering times vary significantly based on scene complexity
- No built-in user session management
MIT License - See LICENSE file for details
This project demonstrates advanced integration of modern AI capabilities with mathematical visualization tools, creating a pipeline from natural language to animated mathematical concepts.