The easiest way to remove vocals from a song for free is Ultimate Vocal Remover (UVR): a free, open-source desktop app that runs AI models like Demucs and MDX on your own computer, with no time limits and no watermark. If you do not want to install anything, LALAL.AI gives you 10 free minutes of AI splitting in the browser, and Audacity can do it offline with Intel’s free OpenVINO plugin.

Below are five free ways to strip vocals (or isolate them for an acapella), with the exact free limits we checked on each tool’s official page in November 2026, step-by-step instructions for the best one, and fixes for the problems most people hit.

Quick answer: the 5 best free vocal removers

# Tool Type Free limit Best for
1 Ultimate Vocal Remover (UVR) 5.6 Desktop app (Windows, macOS, Linux) Unlimited, 100% free and open source (MIT) Best quality, batch jobs, karaoke tracks
2 LALAL.AI Web, desktop and mobile Starter plan: 10 minutes (Relaxed Queue), 200 MB per file One or two songs, no install
3 Audacity + OpenVINO plugin Desktop editor (Windows installer) Unlimited, runs locally Editing the stems right after you split them
4 Moises Web, desktop, iPhone, iPad, Android Limited uploads per month on the free tier Practising an instrument on your phone
5 vocalremover.org Browser Free, processed in your browser A quick test on a simple stereo mix

How we ranked them: separation quality (how much voice “bleeds” into the instrumental), how much you really get for free, privacy (does the file leave your computer?), and how easy the tool is for a beginner. We did not include tools whose free limits we could not confirm on an official page.

How AI vocal removal works (in plain English)

Old karaoke tricks used phase cancellation: most lead vocals sit in the centre of a stereo mix, so if you flip one channel and add it to the other, the centre disappears. It is fast, but it also removes the bass and kick drum (they sit in the centre too) and it fails on mono files.

Modern tools use neural networks trained on thousands of songs where the separate stems are known. Models like Demucs (made by Meta’s research team), MDX-Net and the newer Roformer models “listen” to the full mix and predict which parts of the sound belong to the voice. The result is two or more files: an instrumental (karaoke) track and an acapella, or four stems: vocals, drums, bass and other.

That is why the AI tools below sound far cleaner than any “vocal cut” button in an old audio editor.

1. Ultimate Vocal Remover (UVR): best free vocal remover overall

Ultimate Vocal Remover is the tool most audio hobbyists end up using. Its developers describe it as “100% open-source and free for anyone to use and modify”, and the code is MIT-licensed. The current release is v5.6, with a Windows installer and macOS builds for both Apple Silicon and Intel Macs.

What you get for free: everything. There is no minute counter, no file cap and no watermark, because the AI runs on your own machine. UVR bundles several model families (Demucs v1–v4, MDX-Net and MDX23C, plus Roformer models through its Download Center), so you can pick the model that works best for your song.

Requirements: Windows 10 or later, macOS Big Sur or later, or a Debian/Arch-based Linux (manual install), all 64-bit. It works on a normal CPU, but it is much faster on a graphics card: the project lists an Nvidia GTX 1060 6 GB as the minimum for GPU conversions and recommends 8 GB of VRAM.

How to remove vocals with UVR (step by step)

  1. Download the installer from the Releases page of the official GitHub repository (the file is called UVR_v5.6.0_setup.exe on Windows). Avoid “UVR” downloads on other websites.
  2. Install and open the app. Click Select Input and choose your song (MP3, WAV, FLAC and other common formats work).
  3. Click Select Output and pick a folder for the results.
  4. Under Choose Process Method, pick MDX-Net for a fast, clean vocal/instrumental split, or Demucs if you want four stems (vocals, drums, bass, other).
  5. Choose a model in the dropdown. If a model is not installed yet, open the wrench icon → Download Center and download it first.
  6. Tick GPU Conversion if you have a supported Nvidia card (or an Apple Silicon Mac).
  7. Choose Instrumental Only for a karaoke track, Vocals Only for an acapella, or leave both unticked to get both files.
  8. Press Start Processing. A three-minute song usually takes well under a minute on a GPU and a few minutes on a CPU.

Pros

Cons

2. LALAL.AI: best online vocal remover (no install)

LALAL.AI is a polished stem splitter that works in the browser, on desktop (Windows, macOS, Linux) and on iOS and Android. Its Starter plan is “always free” and includes 10 minutes of processing in the Relaxed Queue, with files up to 200 MB.

One detail matters: LALAL.AI multiplies minutes by the number of stem types you pick. If you split a 3-minute song into vocals only, that uses 3 minutes; splitting it into vocals, drums and bass uses more. So the free plan is best for one to three songs.

Paid plans (if you need more): Lite at $7.50/month billed yearly ($90/year) with unlimited Relaxed Queue and 90 fast minutes, and Pro at $15/month billed yearly ($180/year) with 250 fast minutes, a VST plugin and API access.

How to use LALAL.AI for free

  1. Open lalal.ai and sign in (the free minutes are tied to an account).
  2. Choose Vocal and Instrumental as the stem type.
  3. Drag your song into the upload box and wait for the preview.
  4. Listen to the preview, then process and download the instrumental and the vocal track.

Pros: very clean results, simple interface, works on any device. Cons: only 10 free minutes, the free tier uses the slower queue, and your file is uploaded to their servers.

3. Audacity + Intel OpenVINO plugin: best for editing the stems

Audacity is the free, open-source audio editor many people already have installed (the current version is 4.0.2 for Windows, macOS and Linux). On its own, its old “Vocal Reduction and Isolation” effect uses the phase trick and sounds rough. Intel’s free OpenVINO AI plugins for Audacity change that.

The plugin’s Music Separation effect splits a mono or stereo track into four stems (drums, bass, vocals and other instruments) using Meta’s Demucs v4, and it runs locally on your CPU, GPU or NPU. The plugins are GPL-3.0 licensed. A ready-made installer is offered for Windows; Linux users have to build it from the instructions in the repository.

How to remove vocals in Audacity with the OpenVINO plugin

  1. Install Audacity, then download the latest Windows release of the OpenVINO plugins from Intel’s GitHub page and run it.
  2. Open Audacity → Edit → Preferences → Modules and make sure mod-openvino is enabled, then restart Audacity.
  3. Import your song with File → Import → Audio and select the whole track.
  4. Go to Effect → OpenVINO AI Effects → OpenVINO Music Separation.
  5. Pick the 4-stem option and your device (CPU or GPU), then click Apply.
  6. Audacity adds new tracks for each stem. Mute or delete the vocal track and export the rest with File → Export Audio.

Pros: free and offline, and you can cut, fade and EQ the stems in the same app. Cons: easy installation only on Windows, fewer model choices than UVR, and slower on older CPUs.

4. Moises: best vocal remover app for phones

Moises is built for musicians who want to practise along with a song. It runs on the web, desktop, iPhone, iPad and Android, and it can separate music into up to 27 stem types (on paid plans). Its pricing page says the free tier “provides a limited number of uploads each month, with access to a select set of features”. The exact numbers are only shown after you log in, so check your account before you plan a big batch.

Besides vocal removal, Moises offers tools like key and tempo change and chord detection, which is why guitarists and singers like it.

Pros: excellent mobile apps, practice tools built in. Cons: the free tier is small, and the best stem options are paid.

5. vocalremover.org: quickest browser test

vocalremover.org is a simple page that processes audio directly in your browser, so nothing is uploaded. Its page explains that it uses the classic stereo difference method, and that lead vocals placed in the centre of the mix are removed well. That also means it does not work with one-channel mono files, and it can thin out the bass and drums.

Use it when you just want to hear quickly whether a karaoke version is possible. For anything you will publish or perform with, use UVR or LALAL.AI instead.

Pros: instant, no account, private. Cons: lower quality than AI separation, fails on mono audio.

Tips for a cleaner instrumental

What can you do with the stems?

A note on copyright: removing the vocals does not remove the copyright. Stems from a commercial song are fine for personal practice, but publishing a remix or a karaoke version online usually needs permission from the rights holders, and platforms like YouTube may block or claim it.

FAQ

Can I remove vocals from a song for free without losing quality?

Not perfectly, but very close. AI tools like Ultimate Vocal Remover and LALAL.AI leave far fewer artefacts than older methods. Some faint vocal “ghosting” can remain, especially on reverb-heavy songs, so try a second model or a de-reverb pass if you hear it.

What is the best free vocal remover?

Ultimate Vocal Remover (UVR) is the best free vocal remover for most people: it is open source, has no limits and uses the same model families as many paid services. If you cannot install software, LALAL.AI’s free Starter plan with 10 minutes is the best online option.

How do I remove vocals from a song on my phone?

Use the Moises app (iPhone, iPad, Android) or the LALAL.AI mobile app. Both have free tiers with limits, so they work best for a few songs a month.

Can Audacity remove vocals?

Yes. Audacity’s built-in Vocal Reduction effect works only on centred vocals and sounds rough. For AI-quality results, install Intel’s free OpenVINO plugin and use its Music Separation effect, which splits the song into vocals, drums, bass and other instruments.

Is it legal to remove vocals from a song?

Making a karaoke track for personal use is generally fine. Sharing, selling or uploading the result is different: the instrumental and the acapella are still part of the copyrighted recording, so you usually need permission before publishing them.

Why does the instrumental still have some voice in it?

Reverb, echo, backing harmonies and heavily compressed source files make separation harder. Use the highest-quality source you have, try a different model, and run a de-reverb model afterwards.

Conclusion

For most people the answer is simple: install Ultimate Vocal Remover if you have a computer, and use LALAL.AI’s free 10 minutes when you need a quick result in the browser. Audacity with the OpenVINO plugin is the best pick if you also want to edit the stems, Moises is the best app for phones, and vocalremover.org is fine for a quick test.

Building a website, channel or brand around your music or videos? SSA Agency helps creators and businesses get found on Google and in AI search. Get a free SEO audit and see what is holding your site back.

Leave a Reply

Your email address will not be published. Required fields are marked *