Skip to the page content

Ultimate Vocal Remover (UVR) or a browser tool like RemoveVocal

UVR is the go-to for people who like to tinker: free, open source, packed with model choices, and running entirely on your own machine. RemoveVocal trades that control for a browser tab and no setup. Which is better depends on your computer and your patience.

Updated

What UVR offers

Ultimate Vocal Remover is a free desktop application released under the MIT license. Builds are available for Windows 10 and later, for macOS Big Sur and later on both Apple Silicon and Intel, and installation instructions are provided for Debian- and Arch-based Linux. It bundles several separation approaches, including VR Architecture, MDX-Net and Demucs models, and some Demucs models split music into four stems. It can save results as WAV, FLAC or MP3.

Processing happens on your own computer, so your files do not need to leave it. The project recommends an NVIDIA graphics card with plenty of video memory for fast GPU processing; slower machines can still run it, just with more waiting. Experienced users enjoy chaining models and adjusting settings song by song.

What RemoveVocal offers

RemoveVocal runs the separation on InverseAI's servers. You open the site on a laptop, a phone or a borrowed computer, upload a song or video file, and wait while a progress bar fills. There is one model and nothing to configure. When it finishes, a player lets you switch between Original, Vocals and Instrumental at the same moment, and you download two MP3 files.

That matters most when the device in front of you cannot install software or has no graphics card worth mentioning: a school Chromebook, a work laptop, the tablet you carry to rehearsal.

Where RemoveVocal does less

  • No model choice or settings. You cannot swap models or tune parameters for a hard song.
  • Two stems only. UVR's four-stem models can pull drums and bass apart; RemoveVocal cannot.
  • MP3 only. UVR can write WAV and FLAC.
  • Needs an internet connection and uploads your audio to a server. UVR works locally.
  • Not free without limits. RemoveVocal splits every file of 20 minutes or less for free and plays each result in full in the browser; downloads use account minutes or an Unlimited plan.

Side by side

From UVR's official GitHub repository as of November 2026.
Ultimate Vocal RemoverRemoveVocal
TypeDesktop app you installWebsite
SystemsWindows 10+, macOS Big Sur+, LinuxModern desktop and mobile browsers
ProcessingOn your own computerOn InverseAI's servers
Model optionsSeveral, including VR Architecture, MDX-Net, DemucsOne, no settings
StemsTwo, or four with some Demucs modelsTwo: vocals and instrumental
Output formatsWAV, FLAC, MP3MP3, up to 320 kbps
HardwareNVIDIA GPU recommended for speedAny device with a modern browser
CostFree and open source (MIT)Free processing for songs up to 20 min, full-length playback; minutes or plans for downloads

A simple way to decide

  • You have a gaming PC or a recent Mac, you enjoy experimenting, and you want WAV output or four stems: install UVR.
  • You want the file on your phone before band practice, or you cannot install software on the computer you are using: use RemoveVocal. The page for taking the vocals out of an MP3 is a direct start.
  • You care most about how one particular song sounds: run it through both and compare the same chorus.

Questions people ask

Yes. Its GitHub repository describes it as free and open source under the MIT license, as of November 2026.

No. The separation runs on InverseAI's servers, so any device with a modern browser works.

No. RemoveVocal uses one model, InverseAI's own, with no settings to change.

Your upload and results are stored for a limited time so you can play and download them, then deleted automatically. See the privacy policy for details.

Nothing to install, nothing to configure

Open the page on any device, upload a song, and preview the stems.