gemini-2-podcast Setup Guide

A Python-based tool that generates engaging podcast conversations using Google's Gemini 2.0 Flash Experimental model for script generation and text-to-speech conversion. Now with multi-language support for generating podcasts in various languages.

Features

• Converts content from multiple source formats (PDF, URL, TXT, Markdown) into natural conversational scripts.
• Generates high-quality audio using Google's text-to-speech capabilities.
• Supports multiple languages for podcast generation.
• Provides two distinct voices for dynamic conversations.
• Handles error recovery and retries for robust audio generation.
• Progress tracking with visual feedback during generation.

Prerequisites

Microsoft C++ Build Tools

1. Download Microsoft C++ Build Tools from Visual Studio Installer.
2. Run the installer and select:
   • Desktop development with C++ workload.
   • Optional MSVC build tools (v140, v141, v142) under Installation details.
3. After installation:
   • Reboot your computer.
   • Add MSBuild to system environment variables:

     C:\Program Files (x86)\Microsoft Visual Studio\2019\BuildTools\MSBuild\Current\Bin
     

System Dependencies

For Ubuntu/Debian:

sudo apt-get install ffmpeg portaudio19-dev

For macOS:

brew install ffmpeg portaudio

For Windows:

Install FFmpeg and add it to PATH
PortAudio comes with PyAudio wheels

Project Setup

Clone the Repository:

git clone https://github.com/yourusername/gemini-2-podcast.git
cd gemini-2-podcast

Create and Activate Virtual Environment:

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

Install Python Dependencies:

pip install -r requirements.txt

Create .env File with API Keys:

GOOGLE_API_KEY=your_google_api_key
VOICE_A=Puck
VOICE_B=Kore

Required Files

Ensure these files are present in your project directory:
- generate_podcast.py
- generate_script.py
- generate_audio.py
- system_instructions_script.txt
- system_instructions_audio.txt
- requirements.txt
- README.md

Usage Instructions

Start the Podcast Generation:

Multi-Language Support:

The project supports generating podcasts in multiple languages. Specify the desired language using the --language option. If no language is specified, it defaults to English.

Example usage:

python generate_podcast.py --language spanish

python generate_podcast.py

1. When prompted, input content sources:

   - PDF files: pdf
   - URLs: url
   - Text files: txt
   - Markdown files: md
   

2. Type done when finished.
3. Review the generated script in podcast_script.txt.
4. Press Enter to continue with audio generation or q to quit.

Wait for Audio Generation to Complete:

- A progress bar will display the status.
- Final output: final_podcast.wav.

Output Specifications

- Audio format: WAV
- Channels: Stereo
- Sample rate: 24000Hz
- Bit depth: 16-bit

Contributing

1. Fork the repository.
2. Create a feature branch.
3. Commit your changes.
4. Push to the branch.
5. Open a Pull Request.

License

This project is licensed under the MIT License.

Acknowledgments

• Inspired by NotebookLM's podcast feature.