Podcast Generator

An AI-powered podcast generator that converts PDF documents into natural-sounding conversations.

Demo

Prerequisites

• Docker
• Docker Compose

Setup Instructions

1. Clone the repository:

   git clone git@github.com:anandrmedia/podcastgen.git
   cd podcastgen

2. Create a .env file from the example:

   cp .env.example .env

3. Edit .env and add your OpenAI API key.

4. Configure LLM: In server/src/index.ts, configure the LLM settings according to your needs:

   const llm = new LLM({
       baseUrl: "https://api.openai.com/v1/",    // For OpenAI
       // baseUrl: "https://api.deepseek.com/v1", // For Deepseek
       model: "gpt-4",                          // Model name
       apiKey: process.env.OPENAI_API_KEY       // API key from .env
   });

   Supported LLM providers:

   • OpenAI (api.openai.com)
   • Deepseek (api.deepseek.com)
   • Any compatible API with OpenAI format

5. Create required directories:

   mkdir -p server/src/tmp_data
   mkdir -p server/src/generated-files/scripts
   mkdir -p voices

6. Download Piper voice models:

   • Download the required .onnx voice models from https://huggingface.co/rhasspy/piper-voices/tree/main/en/en_US/
   • Place them in the voices directory Required models (folder structure should exactly be like this):
   • voices/lessac/medium/en_US-lessac-medium.onnx
   • voices/lessac/medium/en_US-lessac-medium.onnx.json
   • voices/kusal/en_US-kusal-medium.onnx
   • voices/kusal/en_US-kusal-medium.onnx.json

7. Start the application:

   docker-compose up

The application will be available at:

• Web UI: http://localhost:3001
• API: http://localhost:3000

Usage

1. Open http://localhost:3001 in your browser
2. Upload a PDF file using the "Upload PDF" button
3. Wait for the processing to complete
4. Select the generated script from the sidebar
5. Click "Play" to start the podcast

Development

To stop the application:

docker-compose down

To rebuild the containers after making changes:

docker-compose up --build

Troubleshooting

If you encounter any issues:

1. Check the Docker logs:

   docker-compose logs

2. Ensure all required voice models are in the voices directory
3. Verify your OpenAI API key is correct
4. Make sure ports 3000 and 3001 are available on your system

License

This project is licensed under the MIT License - see the LICENSE.md file for details.

Third-Party Licenses

This project uses several third-party components, including:

• Piper for text-to-speech conversion
• Piper voice models from rhasspy/piper-voices

For detailed license information of third-party components, see THIRD_PARTY_LICENSES.md.