vocal-remover
A deep-learning tool for separating instrumental and vocal tracks from audio files using PyTorch.
Extract clean stems from audio using neural networks with support for custom model training.
vocal-remover is a Python-based utility designed for audio engineers and researchers who need to isolate vocals from music. It provides a command-line interface to process audio files using pre-trained deep neural networks.
What it is
The tool uses deep learning models to perform source separation, outputting distinct instrumental and vocal WAV files. It includes advanced features like Test-Time Augmentation (TTA) and experimental post-processing to mask instrumentals based on vocal volume.
Who it's for
It suits audio professionals and developers comfortable with Python environments and command-line operations. It is not suitable for users who require a graphical user interface or those without access to a GPU for faster processing.
Key features
Deep Neural Network Separation
Uses advanced U-Net and DenseNet architectures to isolate audio sources.
Test-Time Augmentation
Includes an optional TTA mode to improve the quality of the separation output.
Custom Model Training
Provides a training pipeline to allow users to train models on their own specific datasets.
Experimental Post-processing
Features an optional masking function to refine instrumental tracks based on vocal levels.
Up and running in 3 steps
Our honest verdict
What's great
- Supports GPU acceleration for faster inference and training.
- Includes TTA and post-processing options for higher quality results.
- Allows for custom dataset training to improve model performance.
- Based on established research papers in source separation.
Where it falls short
- Requires manual installation of Python dependencies and PyTorch.
- Command-line only interface with no visual dashboard.
- Experimental post-processing features may cause artifacts.
- High computational cost if training models locally without a powerful GPU.
Pricing & license
The tool is free under its license. The only cost is running it yourself on your own hardware, which requires significant local compute resources for training.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Can I train this on my own data?
Yes, the repository includes a train.py script that allows you to train models using your own directory of instrumental and mixture files.
Does it require a GPU?
It can run on a CPU, but a GPU is highly recommended for reasonable processing speeds, especially when using advanced options like TTA.
