tsurumeso/vocal-remover

Vocal Remover using Deep Neural Networks

What it solves

It provides a way to extract instrumental tracks from songs by separating the vocals from the background music using deep learning.

How it works

The tool uses deep-learning models to process audio files and split them into two distinct tracks: one for vocals and one for instruments. It supports CPU and GPU acceleration and offers advanced options like Test-Time Augmentation (TTA) to improve separation quality and an experimental post-processing mask based on vocal volume.

Who it’s for

Musicians, audio engineers, and anyone looking to create high-quality instrumental versions of existing songs.

Highlights

  • Supports both CPU and GPU inference.
  • Includes Test-Time Augmentation (TTA) for better audio separation.
  • Allows users to train their own custom models using their own datasets.
  • Provides an experimental post-processing feature to refine the instrumental track.

Related

  • Project
  • Project
  • Project
  • Project