Skip to content

deerestFarther/Digital-Life-Project

Repository files navigation

Digital Life Project

Introduction

The Digital Life Project is an innovative endeavor that utilizes cutting-edge technologies in computer vision (CV) to build a digital version of oneself. This project aims to bridge the gap between the physical and digital worlds, offering a unique and personal digital experience.

Demo

The documentations are under revision, you can try the demo at Digital-Life-Project

Features

Developed Features

  • Face Manipulation
    1. Face Swap: Advanced computer vision technology enabling users to swap faces in images or videos, creating a more personalized digital avatar.
    2. Video Retalking: Upcoming feature to manipulate video content so that the digital avatar can mimic user speech, enhancing realism and personalization.
  • Audio Manipulation
    1. Text-to-Speech (TTS): Converts user-input text into natural-sounding speech, enhancing the interactivity of the digital avatar.
    2. Background Music (BGM) Separation: Sophisticated audio processing technology to separate background music from vocals, allowing for clear and distinct audio outputs.

Features in Development

  1. Sound Cloning by Sovits: A future enhancement to clone a user's voice using Sovits technology, providing a more authentic and personalized audio experience for the digital avatar.

  2. LLM-based chatting experience: A future enhancement to chat with a user's voice using Sovits technology, providing a more authentic and personalized chatting experience for the digital avatar.

Getting Started

To get started with the Digital Life Project, follow these steps:

  1. Installation:
  • Basic Installation:

    # Clone the repository
    git clone https://github.com/deerestFarther/Digital-Life-Project.git
    
    # Navigate to the project directory
    cd Digital-Life-Project
    
    # Create and activate a new conda environment
    conda create -n dlp python=3.10
    conda activate dlp
    
    # Install dependencies
    python -m pip install --no-cache-dir -r requirements.txt
    
  • For Nvidia GPU user, you may need to reinstall torch by:

    conda install pytorch torchvision torchaudio pytorch-cuda=11.8 -c pytorch -c nvidia --force-reinstall
    

    This command on works on cuda-11.8 and within conda environment. For more installation command, please check out PyTorch.org

  • For M-series MacOS user, you may need to install pytorch-nightly to use torch on M-SoC MacOS by:

    conda install pytorch torchvision torchaudio -c pytorch-nightly
    
  1. Usage:
  • Starting at local port: 50954, you can change it inside "main.py"
    python main.py
    

Requirements

  • On Win10 with one Nvidia RTX 4090, it requires at least 6G VRAM to use the features mentioned above
  • Still testing on MacOS and cpu-win system

Contributing

We welcome contributions to the Digital Life Project!

License

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

Acknowledgments

Technologies used in this project:

Contributors:

Contact

  • Contact information for the project maintainers.

FAQ

About

If you can't tell, does it matter?

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published