In this guide
Check Audacity and plugin compatibility first. With a compatible OpenVINO installation, use Music Separation and keep the Instrumental track. The old Vocal Reduction and Isolation effect stopped shipping with Audacity 3.5, and channel cancellation is a different, limited method.
To remove vocals in Audacity, use a compatible installation of the OpenVINO Music Separation plugin and keep its Instrumental output. Check compatibility before installing: as of this guide's October 24, 2026 review, Audacity's official plugin download page warns that OpenVINO is not yet compatible with the latest Audacity version. An old tutorial's menu path may therefore be the wrong starting point.
If you are looking for a karaoke backing, you want the music without singing. If you want an acapella, keep the Vocals output instead. Neither is the singer's original studio recording; separation estimates parts from the mix you provide.
Choose a route before installing anything
| Your situation | Starting route | Important check |
|---|---|---|
| You have a compatible Audacity and OpenVINO pair | Music Separation, using the two-stem option | Review Instrumental alone for backing music. |
| An older tutorial shows an effect you cannot find | Check its Audacity version before following its instructions | Vocal Reduction and Isolation is no longer a default effect from 3.5 onward. |
| You want to investigate identical sound in both stereo channels | A center-cancellation test | It can remove wanted instruments too and does not identify singing. |
| You prefer processing an uploaded file without installing a plugin | An online vocal-removal workflow | Check upload limits, the output to keep, and download requirements. |
Keep the original file and a separate copy of your project. Test a short passage containing both an exposed vocal phrase and a busy chorus before committing to a whole-song workflow. A quiet instrumental intro cannot tell you whether the singer will remain in the chorus.
Check OpenVINO compatibility on Windows, Mac or Linux
Start at Audacity's official OpenVINO download page, which offers Windows and macOS links and currently displays the compatibility warning above. Match the plugin's release notes to your installed Audacity, operating system and architecture. Do not infer compatibility from the plugin filename alone or install an arbitrary binary mentioned in a tutorial.
Mac support is not categorically absent: Audacity announced it with 3.7.4 in June 2025. That announcement does not establish compatibility with every later Audacity release. Windows and Linux readers should likewise follow the requirements of the particular package or build instructions they choose, rather than assuming all platforms share one installer.
If no compatible package fits your current setup, keep your working installation. You can wait for compatible support or use a separate workflow for the audio. Changing versions merely to make an old menu appear is not the same as establishing that a plugin supports your setup.
Remove the singing with OpenVINO Music Separation
The steps below follow Intel's Music Separation documentation for a compatible installed plugin. They are a documented workflow, not a report of a separation test on a particular machine.
- Open a copy of the source and select a test passage. Include the difficult vocal overlap you need to evaluate, with enough surrounding music to hear transitions.
- Find OpenVINO Music Separation in the Effect menu. If it is missing, resolve installation and compatibility before continuing.
- Choose the two-stem separation mode. It produces Instrumental and Vocals tracks. The four-stem option instead provides Drums, Bass, Other Instruments and Vocals.
- Choose an available inference device and apply. Allow processing to finish; avoid treating another model or a noise-suppression effect as the same operation.
- Listen to Instrumental by itself. Keep the original and Vocals track out of that comparison. Then check Vocals separately to understand what the separation took away.
Label candidates clearly, for example with the source passage, separation mode and output name. If the first test loses a guitar riff or leaves a chorus lyric, note the source time. A specific failure is more useful than repeatedly processing the previous result and hoping that another pass will repair it.
Why Vocal Reduction and Isolation is missing
Audacity's 3.5 release notes record the removal of Vocal Reduction and Isolation. Its retained manual page describes a separate downloadable Nyquist plugin and the older effect's controls. The historical documentation is not proof that the effect is included in your current installation.
Do not substitute Noise Reduction just because its name sounds similar. A changing singer is a different target from steady background hiss. Decide whether you need estimated music stems or a channel-based experiment, then use the documentation for that method.
Can stereo inversion remove the vocals without AI?
Subtracting one channel from the other cancels samples that match on both sides. It has no knowledge of singers, drums or bass. A centered vocal can be reduced along with other centered material; stereo vocal effects and panned singing can remain. A mono file has no separate left and right information for this experiment.
You can hear the arithmetic in our free center-cancellation tool. Choose its Matching channels sample and Center cancellation: the synthesized matching tones cancel. That demonstrates matching-channel subtraction, not successful vocal removal from a song. Test your own short stereo MP3 or WAV separately and compare lost instruments as well as remaining singing.
If a result becomes thin or nearly silent, turning it up cannot restore the cancelled instruments. Return to the original. The channels versus stems guide explains why extracting a playback channel is different from estimating a musical source.
Export the instrumental without mixing the vocals back in
Before exporting, listen with only the intended instrumental audible. An original mix or separated vocal track playing alongside it can make a good candidate sound as though removal failed. Keep a project copy, exclude unwanted tracks from the export, and check the resulting file in a separate playback step.
Audacity distinguishes saving a project from producing a playable audio file. Its export instructions use File → Export Audio. An .aup3 project is not a replacement for the WAV or MP3 your player needs. A WAV working export also cannot restore detail discarded by an MP3 source.
Review the exported intro, busiest chorus and final ringing note. Check for intelligible lyric remnants, missing backing instruments, abrupt level changes and watery textures. Compare at a similar listening volume; a quieter file can appear cleaner without being a better backing. The vocal-removal artifacts guide helps distinguish these failures, while the karaoke guide covers performance cues and rehearsal checks.
Use an online workflow when installation is not the right fit
The Saturalabs vocal remover accepts MP3, WAV, MP4 or MOV up to 50 MiB. Request “The singing vocals” and inspect Background audio for the remaining arrangement. Isolated sound is the requested singing. Review both before choosing a WAV download, which requires an account and sufficient credits.
This is an alternative workflow, not evidence that it outperforms Audacity on your recording. Start each comparison from the same original passage and judge the sounds you need to retain. If an original instrumental or separate session tracks are available, compare those before accepting an estimated backing.