Podcast Generator

Podcast Generator is a web application that allows users to extract text from a document in PDF or website, generate a podcast conversation outline using Azure OpenAI's GPT-4o (current model), and convert the outline into an audio file using Azure Cognitive Services.

Features

• Extract text from a given PDF or Website
• Generate a podcast conversation outline between two speakers discussing the extracted content.
• Allow users to edit the generated outline before converting it to audio.
• Convert the edited outline to an audio file using Azure Speech Service
• Save the generated audio file with a unique timestamp and allow user to download it

Requirements

• Python 3.9 or higher
• Flask
• Requests
• BeautifulSoup4
• OpenAI Python client
• Azure Cognitive Services Speech SDK

You will need an Azure subscription and create the following resources:

• Azure Blob Storage #To host the audio files
• Azure OpenAI # To generate the scripts of the podcast
• Azure Intelligent Document # to Extract text and context from the pdf files (optional)
• Azure Speech Service #To generates the audio files

Also, your running machine needs to have installed ffmpeg, so to avoid problems, is recommended to run in DevContainers

USE DEV CONTAINER

This repo contains the devContainer config to run it correctly without any inconvenience as a container in your desktop.

Installation

1. Clone the repository:

   git clone https://github.com/frdeange/PodCastSecurityv2.git
   cd PodCastSecurityv2

2a) If you have docker in your computer and WSL available, is recommended to run the folder as container

2b. If not docker or you don't want to use container, enable a virtual environment into your development environment to execute the code

```sh
python3 -m venv venv
source venv/bin/activate
```

3. Install the required packages:

   pip install -r requirements.txt

4. Set up your environment variables for OpenAI and Azure Cognitive Services:

Its recommended to create a .env files on your project root that contains at least the following fields:

AZURE_OPENAI_TYPE="azure" # azure or openAI
AZURE_OPENAI_KEY="YOUR_OPEN_AI_KEY"
AZURE_OPENAI_ENDPOINT="YOUR_OPENAI_ENDPOINT"
AZURE_OPENAI_API_VERSION="YOUR_API_VERSION" 
AZURE_OPENAI_ENGINE="gpt-4o" # Like gpt-4o or model you deployed

#Azure Intelligent Document Required Params
AZURE_FORM_RECOGNIZER_KEY="YOUR INTELLIGENT DOCUMENT KEY"
AZURE_FORM_RECOGNIZER_ENDPOINT="YOUR INTELLIGENT DOCUMENT ENDPOINT"

#Azure Speech Required Params
AZURE_SPEECH_KEY="YOUR SPEECH SERVICE KEY"  
AZURE_SPEECH_REGION="YOUR SPEECH SERVICE region" # For example swedencentral 

#Azure Storage Required Params
AZURE_STORAGE_CONNECTION_STRING="YOUR CONNECTION STREAM"
AZURE_STORAGE_CONTAINER_NAME="YOUR BLOB CONTAINER NAME"

Usage

1. Run the Flask application:

   python app.py

2. Open your web browser and go to http://127.0.0.1:5000.

3. Use the web interface to:

   • Extract text from a website by providing the URL or uploading a .pdf file
   • Generate a podcast conversation outline.
   • Edit the generated outline if necessary.
   • Convert the outline to an audio file.

API Endpoints

• POST /extract_text_from_website: Extract text from a given website URL.

  • Request: website-url (form data)
  • Response: JSON with extracted text

• POST /generate_outline: Generate a podcast conversation outline.

  • Request: content (form data)
  • Response: JSON with generated outline

• POST /generate_audio: Convert the edited outline to an audio file.

  • Request: ssml_content (form data)
  • Response: JSON with the path to the generated audio file

• GET /audio/<filename>: Retrieve the generated audio file.

  • Request: filename (URL parameter)
  • Response: Audio file

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.

License

This project is licensed under the MIT License. See the LICENSE file for details.