AI cover songs

Imagine taking your favorite song and making it sound like it's sung by someone else - that's what this does! You can use it to:

  • Turn any YouTube song into an AI cover
  • Make local audio files sing in a different voice
  • Add singing abilities to your AI projects

What Can It Do?

  • Easy-to-use web interface for making covers
  • Take songs from YouTube or your computer
  • Save the individual parts of songs (vocals, background music)
  • Control how the voice sounds (pitch, volume, echo effects)
  • Switch between different AI voices
  • Save final songs as MP3 or WAV files

Technical Features

Core Features

  • RVC v2 voice model compatibility
  • RMVPE pitch extraction engine
  • Web-based user interface
  • Local file processing support
  • Pixeldrain integration for model downloads

Audio Controls

  • Main/backup vocals volume adjustment
  • Instrumental volume control
  • Index Rate modification (0-1)
  • Reverb parameters: Size (0-1)Wetness (0-1)Dryness (0-1)Damping (0-1)
  • Size (0-1)
  • Wetness (0-1)
  • Dryness (0-1)
  • Damping (0-1)

Voice Processing

  • Pitch detection algorithms: rmvpe/mangio-crepe
  • Filter radius: 0-7 (default: 3)
  • RMS mix rate: 0-1 (default: 0.25)
  • Voice protection: 0-0.5 (default: 0.33)

How to Install

What You Need First

  • Git - for downloading the project
  • Python 3.9 - must be version 3.9, other versions might not work
  • ffmpeg - for handling audio files
  • sox - for sound processing

Installation Steps

  • Clone the project & install dependencies:
pip install -r requirements.txt
  • Get the required AI models:
python src/download_models.py

How to Use It

Easy Way (Web Interface)

Just run this command:

bashCopypython src/webui.py

Extra options for web interface:

  • --share: Share with others online
  • --listen: Let others on your network use it
  • --listen-host: Choose which network address to use
  • --listen-port: Choose which port to use

Advanced Way (Command Line)

Basic command:

bashCopypython src/main.py -i SONG_INPUT -dir RVC_DIRNAME -p PITCH_CHANGE [options]

Must-have settings:

  • -i: Your song (YouTube link or file path)
  • -dir: Which voice to use
  • -p: How much to change the pitch (0 = no change)

Optional settings:

  • -k: Keep all the separate audio files
  • -ir: How much accent to keep (0-1)
  • -fr: How smooth the voice should be (0-7)
  • -rms: Voice loudness control (0-1)
  • -mv: Main voice volume
  • -bv: Background voice volume
  • -iv: Music volume
  • -oformat: Choose MP3 or WAV

Where Files Go

Copy├── rvc_models          (Voice files go here)
├── mdxnet_models       (AI models live here)
├── song_output         (Finished songs go here)
└── src                 (Program files)

Rules for Using

Please Don't:

  • Use to attack or make fun of people
  • Spread political or religious messages
  • Share inappropriate content
  • Sell the voices or songs
  • Pretend to be someone else to cause harm
  • Use for scams or fraud
↗ X ↗ LinkedIn