Ultimate Vocal Remover v5 Review

Ultimate Vocal Remover is a desktop GUI for AI source separation: load audio files, pick a model, and it writes out separated vocals and instrumentals (or 4-stem drums, bass, other and vocals with Demucs models). Processing happens on your own computer, so nothing is uploaded. The current stable release is v5.6 (September 2023), which supports VR Architecture, MDX-Net, MDX23C and Demucs v1 to v4 models, ensembling of several models, and a built-in Download Center for additional models and patches. A beta build adds Roformer, SCNet and Bandit models and, on Windows, DirectML for AMD Radeon and Intel Arc GPUs. The project README says UVR's core developers trained all bundled models except the Demucs v3 and v4 4-stem models.

The Windows installer bundles Python, PyTorch and dependencies (Windows 10 or later, and it must be installed to the C: drive). macOS builds exist for Apple Silicon and Intel (Big Sur and later), with MPS GPU acceleration for Demucs v4 and MDX-Net; first launch can take 5 to 10 minutes, and unsigned-app security prompts may need Terminal commands. Linux users install from source (Debian and Arch instructions). On October 6, 2026 the lead developer posted that UVR 6 is coming "within the next six weeks", rebuilt on PySide with support for newer models and better Linux packaging.

There are no paid plans, credits or usage limits on the official site; the developers ask for optional donations through Buy Me a Coffee. Your real cost is hardware and time: the README lists an Nvidia GTX 1060 6GB as the minimum for GPU conversions and recommends Nvidia GPUs with at least 8 GB of VRAM; CPU-only processing works but is slower.

Beware of look-alike sites selling "UVR online" subscriptions: the official site links only to the GitHub release page for downloads.

Prices checked on the vendor's site on November 3, 2026. See ultimatevocalremover.com and the v5.6 release page.

What the UVR site, README and release notes list:

UVR suits producers, remixers, karaoke makers and hobbyists who separate a lot of tracks, have a reasonably modern computer, and want to avoid per-minute fees or uploading unreleased audio. Its model choice and ensembling give more control than most web tools, at the cost of a technical interface, large downloads and a development pace that slowed after 2023 (UVR 6 is announced but not released at the time of writing).

If you want separation in a browser or on a phone with no setup, try LALAL.AI or Moises. Musicians who also want chord detection, tempo tools and practice features should look at Moises.

UVR is widely used and discussed: the GitHub repository shows about 27,000 stars and over 1,500 open issues, many of them install and model-compatibility reports. A review by Aiseesoft, which sells a competing vocal remover, credits UVR with being free, private and offering several model families, and lists a less polished interface, bleed on live or dense mixes, and slow CPU-only processing as drawbacks. A RouteNote walkthrough calls UVR5 the closest contender to clean separation it has found, while noting it is not the most intuitive to use and models must be downloaded one at a time. The lead developer's own October 2026 update acknowledges that GitHub has been quiet since 2023 and that he is the only active developer on the main codebase.