title	emoji	colorFrom	colorTo	sdk	sdk_version	app_file	pinned
AI Lecture Forge	📚	blue	green	gradio	4.0.0	src/app.py	false

AI Lecture Forge

🏆 Production-ready AI system for transforming conversational transcripts into structured teaching material.

Overview

AI Lecture Forge is a state-of-the-art system that leverages multiple AI models to transform raw transcripts into professional teaching materials. It supports both local inference and API-based models, with optional Text-to-Speech capabilities.

Key Features

Multiple AI Models

• Local Models (No API key required):
  • Microsoft Phi-4 (default)
  • NovaSky Sky-T1-32B
  • DeepSeek V3
• OpenAI Models (API key required):
  • Default: gpt-4o-mini
  • Supports all OpenAI models
• Google Models (API key required):
  • Default: gemini-2.0-flash-exp
  • Supports all Gemini models

Text-to-Speech (Optional)

• Kokoro-82M voice synthesis
• High-quality audio generation
• Multiple voice options

Core Features

• PDF transcript processing
• Customizable lecture duration (30-60 minutes)
• Practical examples integration
• Structured output with sections
• Real-time processing

Technologies Used

• Python 3.8+
• PyTorch + Transformers: Local inference with multiple models
  • Microsoft Phi-4 (default)
  • NovaSky Sky-T1-32B
  • DeepSeek V3
• Text-to-Speech: Kokoro-82M for voice generation
• Gradio 4.0.0: Web interface framework
• OpenAI API (optional): Multiple models
  • Default: gpt-4o-mini
  • Supports all OpenAI models
• Google Gemini API (optional): Multiple models
  • Default: gemini-2.0-flash-exp
  • Supports all Gemini models
• PyPDF2: PDF processing
• python-dotenv: Environment management
• tiktoken: Token counting for OpenAI
• tqdm: Progress bars
• numpy: Numerical operations

Quick Start

Local Development

# Clone repository
git clone [repository-url]
cd ai-lecture-forge

# Install dependencies
pip install -r requirements.txt

# Optional: Configure API keys
# Create .env file if using API models
touch .env
echo "OPENAI_API_KEY=your_key" >> .env
echo "GOOGLE_API_KEY=your_key" >> .env

# Run application
python src/app.py

Hugging Face Spaces Deployment

1. Create new Space
   • Select SDK: Gradio
   • Hardware: T4 GPU (recommended)
2. Connect repository
3. Optional: Add API keys in Settings
   • OPENAI_API_KEY (if using OpenAI models)
   • GOOGLE_API_KEY (if using Gemini models)

API Keys Configuration

There are two ways to provide API keys:

1. Through Web Interface (Recommended)

When using API models, the interface will show secure input fields for:

• OpenAI API Key (when using OpenAI models)
• Gemini API Key (when using Gemini models)

This is the recommended method for:

• Public deployments
• Hugging Face Spaces
• Shared instances

2. Using Environment Variables

For local development or private deployments:

# Optional: Create .env file
OPENAI_API_KEY=your_openai_key
GOOGLE_API_KEY=your_gemini_key

Note: If both methods are used, keys provided through the interface take precedence.

Environment Configuration

The .env file is optional and only required when using API-based models:

# Required for OpenAI models
OPENAI_API_KEY=your_openai_key

# Required for Gemini models
GOOGLE_API_KEY=your_gemini_key

Note: Local models (Phi-4, Sky-T1-32B, DeepSeek V3) work without any API keys.

Model Selection Guide

Local Models

• Best for: Development, testing, offline use
• No API costs
• GPU recommended for better performance

API Models

• Best for: Production, high-quality output
• Requires API keys
• Pay-per-use pricing

TTS Integration

• Optional feature
• Adds voice synthesis capability
• Uses Kokoro-82M model

Performance Considerations

• GPU recommended for local models
• T4 GPU or better for optimal performance
• CPU-only mode available but slower
• API models not affected by hardware

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT License - See LICENSE for details.

Competition & Production Use

This project is designed for production environments and competitive scenarios:

• Production-ready error handling
• Scalable architecture
• Multiple model fallbacks
• Comprehensive logging
• Performance optimization