NeuralNoise: The AI Podcast Studio

NeuralNoise is an AI-powered podcast studio that uses multiple AI agents working together. These agents collaborate to analyze content, write scripts, and generate audio, creating high-quality podcast content with minimal human input. The team generates a script that the cast team (using a TTS tool of your choice) will then record.

Examples

Source	Type	NeuralNoise	Download
TikTok owner sacks intern for sabotaging AI project	🌐 Web article	example_neuralnoise_0.mp4	Link
Before you buy a domain name, first check to see if it's haunted	🌐 Web article	example_neuralnoise_1.mp4	Link
Linus Torvalds Comments On The Russian Linux Maintainers Being Delisted	🌐 Web article	example_neuralnoise_2.mp4	Link
Hallo2: Long-Duration and High-Resolution Audio-Driven Portrait Image Animation	📗 PDF	example_neuralnoise_3.mp4	Link
Ep17. Welcome Jensen Huang \| BG2 w/ Bill Gurley & Brad Gerstner	📺 YouTube	example_neuralnoise_4.mp4	Link

Objective

The main objective of NeuralNoise is to create a Python package that simplifies the process of generating AI podcasts. It utilizes OpenAI for content analysis and script generation, ElevenLabs for high-quality text-to-speech conversion, and Streamlit for an intuitive user interface.

Features

Content analysis and script generation using OpenAI's language models
High-quality voice synthesis with ElevenLabs or OpenAI
Audio processing and manipulation with pydub
Easy way to modify parts of the script and regenerate the podcast

Installation

To install NeuralNoise, follow these steps:

Install the package:

pip install neuralnoise

or from source:

git clone https://github.com/leopiney/neuralnoise.git
cd neuralnoise
pip install .

Set up your API keys:
- Create a .env file in the project root
- Add your OpenAI and ElevenLabs API keys:
```
OPENAI_API_KEY=your_openai_api_key

# Optional
ELEVENLABS_API_KEY=your_elevenlabs_api_key
```

Usage

To run the NeuralNoise application first make sure that you create a configuration file you want to use. There are examples in the config folder.

Then you can run the application with:

nn new <url|filepath> --name <name> --config-path <config>

Want to edit the generated script?

The generated script and audio segments are saved in the output/<name> folder. To edit the script:

Locate the JSON file in this folder containing all script segments and their text content.
Make your desired changes to specific segments in the JSON file. Locate the "sections" and "segments" content in this file that you want to change, then feel free to edit the content of the segments you want to change.
Run the same command as before (same name) to regenerate the podcast.

The application will regenerate the podcast, preserving unmodified segments and only processing the changed ones. This approach allows for efficient editing without regenerating the entire podcast from scratch.

Roadmap

Better PDF and articles content extraction.
Add local LLM provider. More generic LLM configuration. Leverage AutoGen for this.
Add local TTS provider
Add podcast generation format options: interview, narrative, etc.
Add podcast generation from multiple source files
Add more agent roles to the studio. For example, a "Content Curator" or "Content Researcher" that uses tools to find and curate content before being analyzed. Or a "Sponsor" agent that adds segways to ads in the podcast script (à la LTT).
Add music and sound effects options
Real-time podcast generation with human and AI collaboration (🤔)

Contributing

Contributions to NeuralNoise are welcome! Please feel free to submit a Pull Request.

License

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

bubudrc/neuralnoise