Ultimate Vocal Remover (UVR) is a desktop application that uses deep neural networks to separate a song into its vocal and instrumental stems. Its GUI wraps several state-of-the-art source-separation architectures, including MDX-Net, Demucs and VR models, and is widely used to create karaoke tracks, acapellas, remixes and clean training data for other audio models. The project has attracted roughly 26,700 stars on GitHub and remains the most popular open-source tool in its category.
Source separation used to require expensive studio software or command-line research code. UVR packages the best open separation models behind a point-and-click interface: choose an audio file, pick a model or an ensemble of models, and export separated stems in common audio formats. A built-in download center keeps the model library current, ensemble modes combine multiple architectures for higher-quality results, and GPU acceleration is supported where available. Ready-made installers for Windows and macOS mean musicians and producers can use it without touching Python, while the underlying project stays fully open for developers to inspect and extend.
Ultimate Vocal Remover is completely free and open source under the MIT license. There are no subscriptions or per-track charges; optional community support channels exist for users who want to contribute.
UVR is for musicians, DJs, producers, karaoke creators and content makers who need clean vocal or instrumental stems, and for audio researchers who want a convenient front end for modern source-separation models.