Ultimate Vocal Remover

Ultimate Vocal Remover wraps the strongest open stem-separation models in a desktop application that anyone can install and run. Rather than betting on one architecture, it aggregates MDX-Net (including the newer MDX23C models), Demucs v3 and v4 four-stem variants, and the VR architecture, letting users pick models per track or chain them in ensembles, with batch processing and detailed settings management built in. Packaged installers cover 64-bit Windows 10 and later and macOS Big Sur and later in both Apple Silicon and Intel builds, while Linux users install from source; version 5.6 is current. Processing runs on CPU, but the project recommends an NVIDIA GPU, with a GTX 1060 6 GB stated as the minimum for GPU conversions and 8 GB or more of VRAM preferred, and AMD and Intel Arc cards get partial support through a separate DirectML build. The GUI code is MIT licensed and the repository has about 25,600 stars. Musicians making instrumentals and karaoke tracks, AI cover creators, and dataset builders rely on its model zoo daily.