Local-NotebookLM

A local AI-powered tool that converts PDF documents into engaging podcasts, using local LLMs and TTS models.

Features

• PDF text extraction and processing
• Customizable podcast generation with different styles and lengths
• Local LLM support through various providers (LMStudio, Ollama, etc.)
• Text-to-Speech conversion with voice cloning capabilities
• Fully configurable pipeline

Installation

1: Clone the repository:

git clone https://github.com/yourusername/Local-NotebookLM.git
cd Local-NotebookLM

2: Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows, use: venv\Scripts\activate

3: Install the required packages:

pip install -r requirements.txt

Configuration

Create a config.yaml file in the root directory with the following structure:

Global:
  output_dir: "./resources"
  provider_format: 'openai'  # Options: ['openai', 'mlx_lm']
  provider: "lmstudio"       # Options: ['openai', 'lmstudio', 'ollama', 'groq', 'other']
  base_url: ''              # Required only if provider is 'other'
  api_key: ''               # Required for 'other', 'openai', and 'groq' providers

Step1:
  model_name: "mlx-community/Josiefied-Qwen2.5-1.5B-Instruct-abliterated-v1-4bit"
  max_tokens: 512
  temperature: 0.7
  chunk_size: 1000
  max_chars: 100000

Step2:
  model_name: "mlx-community/Josiefied-Qwen2.5-14B-Instruct-abliterated-v4-4-bit"
  max_tokens: 8126
  temperature: 1
  length: "long"           # Options: ["short", "medium", "long", "very-long"]
  style: "academic"        # Options: ["friendly", "professional", "academic", "casual", "technical", "funny"]

Step3:
  model_name: "mlx-community/Josiefied-Qwen2.5-14B-Instruct-abliterated-v4-4-bit"
  max_tokens: 8126
  temperature: 1

Step4:
  model_name: "lucasnewman/f5-tts-mlx"
  cohost_speaker_ref_audio_path: "./voices/cohost.wav"
  cohost_speaker_ref_audio_text: "Some call me nature, others call me mother nature."

Usage

1. Prepare your environment:

   • If using LMStudio: Start LMStudio and ensure the API server is running
   • If using Ollama: Install and start Ollama with your desired models
   • If using other providers: Ensure you have the necessary API keys

2. Run the script:

python main.py input.pdf [options]

Available options:

• --config_path: Path to custom config file (default: config.yaml)
• --output_dir: Directory for output files
• --length: Desired podcast length (short/medium/long/very-long)
• --style: Podcast style (friendly/professional/academic/casual/technical/funny)
• --chunk_size: Size of text chunks for processing
• --max_chars: Maximum characters to process from PDF

Example:

python main.py research_paper.pdf --style academic --length long

Pipeline Steps

1. PDF Processing (Step1)

   • Extracts text from PDF
   • Cleans and formats the content
   • Splits into manageable chunks

2. Transcript Generation (Step2)

   • Generates initial podcast script
   • Applies specified style and length
   • Structures content for audio format

3. TTS Optimization (Step3)

   • Rewrites content for better TTS performance
   • Adds speech markers and formatting
   • Optimizes for natural flow

4. Audio Generation (Step4)

   • Converts text to speech
   • Applies voice cloning if specified
   • Generates final audio file

Requirements

• Python 3.8+
• MLX (for local AI models)
• PDF processing libraries
• TTS dependencies
• Additional requirements listed in requirements.txt